Cellular Production Lines with Asymptotically Reliable Bernoulli Machines: Lead Time Analysis and Control

Size: px
Start display at page:

Download "Cellular Production Lines with Asymptotically Reliable Bernoulli Machines: Lead Time Analysis and Control"

Transcription

1 Cellular Production Lines with Asymptotically Reliable Bernoulli Machines: Lead Time Analysis and Control Semyon M. Meerkov and Chao-Bo Yan Department of Electrical Engineering and Computer Science University of Michigan, Ann Arbor, MI Abstract Cellular lines are production systems consisting of cells comprised of machines performing similar operations. These systems are notorious for having excessively long lead time (LT), often orders of magnitude longer than the total processing time by all machines in the system. The main goal of this paper is to provide a plausible explanation of this phenomenon and offer methods for its alleviation. To accomplish this, the paper develops a technique for performance evaluation of cellular lines with asymptotically reliable Bernoulli machines and uses it to analyze and control LT. It shows, in particular, that robustness of LT is a decreasing function of the number of machines in a cell. Thus, in cells with many machines, small variations in either release rates or machine efficiencies may lead to dramatic increases of LT, if operating points are not selected appropriately. A method for selecting appropriate operating points and, thus, controlling release rates in the open-loop regime is provided. In addition, feedback control of raw material release is considered and shown to be effective for improving lead time performance for any operating point. Keywords: Production systems, Cellular lines, Lead time, Asymptotically reliable Bernoulli machines, Open- and closed-loop release control. This work was supported by NSF Grant No. CMMI

2 Introduction This paper considers production lines consisting of cells comprised of machines performing similar operations. Such a production line with M cells, c, c,..., c M, each comprised of C i, i =,,..., M, machines and M buffers, b, b,..., b M, is shown in Figure.. These lines are referred to as cellular. Cellular lines are often encountered in machining operations (e.g., turning, boring, milling, drilling, and grinding cells in automotive transmission plants). Under a fixed dispatch policy, re-entrant lines in semiconductor manufacturing also can be viewed as cellular (e.g., lithography, etching, and deposition tool groups). Cellular lines are sometimes referred to as series-parallel production systems. Figure.: Cellular production line In practice, performance of cellular lines is often marred by excessively long lead time (LT), i.e., the average time a part spends in the system, being processed and waiting for processing. In some cases, LT has been observed to be orders of magnitude longer than the total processing time by all machines in the system. (As an anecdotal evidence, the first author encountered gears in a machining department of an automotive transmission plant, which rusted through while waiting for processing.) Therefore, performance analysis of cellular lines and, in particular, of LT and, based on this analysis, insights and recommendations for management and control are important theoretical and practical problems. Unlike production systems with a single machine per stage, which have been analyzed extensively in the literature (see, for instance, monographs []-[8]), quantitative analysis of cellular lines received relatively little attention. Related articles can be classified into two groups: those that aggregate parallel machines into a single machine (see [9]-[6]) and those that develop exact methods for performance analysis of two-cell systems (see [7]-[]). The aggregation approach provides

3 a relatively low accuracy of performance measures. The exact methods are typically computationally intensive. Also, while [9]-[] present methods for performance evaluation, none considers raw material release mechanisms and the resulting LT as a function of system parameters. In our previous work [], we introduced the problem of LT analysis and control for the usual, serial lines. In the current paper, we investigate the cellular systems. The development is based on several simplifying assumptions. First, the machines are assumed to obey the Bernoulli reliability model. According to this model, a machine is up during a cycle time with probability p and down with probability p. This implies that the number of machines in a cell, which are up during a cycle time, is distributed according to a binomial distribution and, therefore, such lines may also be called binomial. The parameter p is referred to as the machine efficiency. The second assumption postulates that the machines are asymptotically reliable, i.e., p i j = εk i j, i =,,..., M, j =,,..., C i, (.) where p i j is the efficiency of machine m i j (i.e., the j-th machine in the i-th cell), 0 < ε is a small parameter, and k i j is a number of order one. The third assumption, introduced to make the results more transparent, postulates that the number of machines in each cell is the same, i.e., C i = C, i =,,..., M, (.) and, in addition, the machines in a cell are of equal efficiency, i.e., p i j = p i, i =,,..., M, j =,,..., C. (.3) This implies, of course, that k i j = k i, j. Next, it is assumed that the capacity, N i, of all buffers, b, b,..., b M, is infinite, i.e., N i =, i =,,..., M. (.4)

4 Finally, the last assumption refers to the release mechanism of raw material. We model the release mechanism by another cell, c 0, also consisting of C identical asymptotically reliable Bernoulli machines, each defined by p 0 = εk 0, (.5) where, as before, 0 < ε and k 0 is of order one. While the efficiencies, p i, i =,,..., M, of the producing machines are fixed, the efficiency, p 0, of the release machines, is free and viewed as a design parameter, to be selected so that the lead time, LT, takes the desired value. The value of p 0 is called the raw material release rate. A cellular line with such a release mechanism is shown in Figure., where the release cell, c 0, and the raw material buffer, b 0, are indicated in gray. Figure.: Cellular line with a raw material release cell Under these assumptions, this paper develops a method for performance evaluation of cellular lines, investigates structural and quantitative properties of LT, and offers open- and closed-loop methods for LT control by selecting an appropriate p 0 or an equivalent hourly release rate. More precisely, the paper presents the following novel results: An analytical method for performance evaluation of cellular lines with asymptotically reliable Bernoulli machines. An expression for LT as a function of the producing machine efficiency and raw material release rate. An expression for LT as a function of the so-called relative load on the system, ρ = p 0, min p i i M 3

5 and shows that this function has a knee, beyond which LT grows extremely fast; the position of this knee is quantified analytically. Robustness properties of LT with respect to p 0 (i.e., when the release rate is not exactly as expected) and with respect to p i (i.e., when the producing machine efficiency is not exactly the one nominally assumed). An expression for the release rate that ensures the desired LT, while maximizing the production rate. A method for calculating deterministic, hourly, raw material release rate, which results in the performance being close to the desired LT. A feedback control law for raw material release, which ensures the desired LT even when the operating point is beyond the knee and the machine efficiency is different from nominally assumed. In spite of the simplifying assumptions, under which these results are obtained, we believe they provide useful insights into the lead time behavior of cellular lines. The outline of this paper is as follows: In Section, a formal model of a cellular line with asymptotically reliable Bernoulli machines and a raw material release mechanism is introduced. Section 3 develops a method for its performance evaluation. Structural and quantitative analyses of LT in systems with identical machines are carried out in Section 4. Some of these results are extended to systems with non-identical machines in Section 5. Deterministic, hourly, release rate is investigated in Section 6, and feedback control of LT is discussed in Section 7. The conclusions and topic for future work are given in Section 8. All proofs are included in the Appendices. Modeling and Problems Addressed The following model is considered throughout this paper: (i) The system consists of M producing cells, c, c,..., c M ; a release cell, c 0 ; M work-inprocess buffers, b, b,..., b M ; and a raw material buffer, b 0 (see Figure.). 4

6 (ii) Each cell c i, i = 0,,..., M, is comprised of C machines, m i, m i,..., m ic. (iii) Each machine, m i j, i = 0,,..., M, j =,,..., C, is asymptotically reliable and obeys the Bernoulli reliability model with the efficiency p i j = p i = εk i, (.) where 0 < ε and k i is a number of order one. All machines have identical cycle time, τ. The time is slotted with the slot duration τ. (iv) All buffers are of infinite capacity. (v) Machines in the release cell, c 0, are never starved. Machines in the producing cells, c, c,..., c M, may be either starved or non-starved according to the following convention: Let h i (n), i = 0,,..., M, be the number of parts in buffer b i at the beginning of time slot n {,,...} and C i+ (n) be the number of machines in cell c i+ that are up during the time slot n. Then, if h i (n) C i+ (n), none of the machines in cell c i+ is starved; if 0 < h i (n) < C i+ (n), C i+ (n) h i (n) machines are starved; if h i (n) = 0, the whole cell c i+ is starved. Note: In general, raw material release mechanism could be modeled as a single machine producing parts according to a binomial distribution. However, we find it more convenient to model the release as adopted in this paper, i.e., as a cell with Bernoulli machines. Nevertheless, considering hourly release (see Sections 6 and 7), we remove this assumption and provide recommendations for releasing a deterministic amount of raw material, which leads to LT being close to the desired. Given the model defined by assumptions (i)-(v), the problems addressed in this paper are: Develop a method for evaluating the production rate (PR), work-in-process (WIP), and lead time (LT). Analyze LT as a function of system parameters and, in particular, investigate robustness of LT with respect to p 0 and p i. 5

7 Based on these analyses, provide a plausible explanation for poor lead time performance of cellular lines. Offer a method for calculating raw material release rates that ensure the desired LT, while maximizing PR. Develop a feedback control policy for raw material release that leads to the desired LT. Solutions of these problems are given in Sections Performance Evaluation In this section, we develop a method for performance evaluation of cellular lines defined by assumptions (i)-(v). We begin with the simplest case of a single producing cell and a release cell and then extend the results to multiple producing cells. 3. Single producing cell Theorem 3. Under assumptions (i)-(v) with M = and p 0 < p, the production rate of the cellular line is given by PR = Cp 0 ; (3.) there exists ε 0 such that for all 0 < ε ε 0, the expressions ŴIP = C + pc 0 ( pc 0 ), (3.) p C p C 0 LT = C Cp 0 + pc 0 ( p C 0 ) C(p C p C 0 ) (3.3) 6

8 provide estimates of WIP and LT with the accuracy O(ε ), i.e., ŴIP WIP = O(ε ), LT LT = O(ε ), (3.4) where LT and LT are in units of the machine cycle time. Proof: See the Appendix. Note that when C =, estimates (3.) and (3.3) reduce to the exact values of WIP and LT derived in [] for the usual serial lines. Also, as can be surmised from the proof of Theorem 3., estimates of WIP and LT can be derived with accuracy higher than O(ε ) by keeping additional terms in Taylor expansions of the transition probabilities. It is of interest to evaluate ε 0, which would result in practically acceptable accuracy of estimates (3.) and (3.3). This could be carried out analytically, but the results are typically very conservative. Therefore, we carry this out by simulating the systems at hand, evaluating WIP and LT, and comparing them with ŴIP and LT. Each simulation runs for 00, 000 time slots, including 50, 000 time slots of warm-up period; 0 simulation runs are carried out for each set of system parameters. The accuracy of ŴIP and LT is evaluated by ɛ WIP = ŴIP WIP WIP 00% and ɛ LT = LT LT LT 00%. The results presented in Table 3. show that the accuracy is quite high for p [0.9, 0.99] and ρ [0.85, 0.99], where ρ = p 0 p. Thus, we conclude that ε 0 can be selected as Multiple producing cells For performance analysis of cellular lines with M > producing cells, we use a generalization of the recursive aggregation procedure developed in [7] for serial (non-cellular) production lines with Bernoulli machines. According to this procedure, pairs of consecutive cells are recursively aggregated into a single cell, using the so-called backward and forward aggregations. As a result, the cellular line with M producing cells and a release cell is represented by M cellular lines, each comprised of two cells with aggregated machines. Then, using expressions (3.) and (3.3) with p replaced by the efficiency of the aggregated machines (which, for the model (i)-(v), turn out to be p, p,..., p M ), we introduce the following estimates of WIP i and LT for cellular lines defined by 7

9 Table 3.: Accuracy of estimates (3.) and (3.3) C p ρ Simulation Estimate Error WIP LT ŴIP LT ɛ WIP ɛ LT %.33% % 0.48% %.74% %.5% %.53% % 0.0% %.% %.40% % % 0.75% % 8.8% %.94% %.48% %.3% % 6.63% % 3.0% % 4.% % 5.0% %.3% % % 0.55% assumptions (i)-(v) with M > cells: ŴIP i = C + pc 0 ( pc 0 ) p C i+, i = 0,,..., M, (3.5) pc 0 LT = M C Cp 0 + M i= p C 0 ( p C 0 ) C(p C i p C 0 ), (3.6) where, as before, LT is in units of the machine cycle time. As for the production rate, PR, of such lines, its exact value remains the same as in the case of M = (see the Appendix), i.e., PR = Cp 0. (3.7) The accuracy of estimates (3.5) and (3.6) has been evaluated numerically, using the simulation 8

10 procedure described in Subsection 3.. The results for cellular lines with ten cells having identical and non-identical machines are shown in Tables 3. (with ρ = p 0 p ) and 3.3 (with ρ max = respectively. Lines L -L 3 involved in Table 3.3 are given by p 0 min i M p i ), L : p = [0.99, 0.93, 0.96, 0.95, 0.99, 0.97, 0.95, 0.9, 0.98, 0.95]; L : p = [0.9, 0.94, 0.98, 0.9, 0.9, 0.9, 0.9, 0.96, 0.93, 0.9]; (3.8) L 3 : p = [0.96, 0.98, 0.99, 0.97, 0.9, 0.94, 0.99, 0.99, 0.94, 0.99], where the p i s have been selected randomly and equiprobably from the interval [0.9, 0.99]. As it follows from these tables, expressions (3.5) and (3.6) provide relatively precise estimates of WIP and LT when ε 0 = 0.. Table 3.: Accuracy of estimates (3.5) and (3.6) (identical machine case) C p ρ Simulation Estimate Error WIP LT ŴIP LT ɛ WIP ɛ LT %.3% % 0.39% %.9% % 4.43% %.0% % 0.73% % 0.34% %.76% %.49% % 0.69% % 0.9% % 0.36% % 9.38% % 4.07% %.4% %.3% % 7.37% % 3.74% % 0.47% % 3.50% % 5.0% %.4% % 0.74% % 0.8% 9

11 Table 3.3: Accuracy of estimates (3.5) and (3.6) (non-identical machine case) C line ρ max Simulation Estimate Error WIP LT ŴIP LT ɛ WIP ɛ LT 5 L %.57% % 0.09% % 3.04% %.% L %.8% % 0.58% %.60% %.05% L %.5% % 0.46% %.8% % 0.84% L % 6.5% %.68% % 0.7% %.87% L % 5.33% %.0% % 9.73% % 4.8% L % 5.63% %.43% % 9.5% %.% Based on the results of this section, we conclude that (3.5) and (3.6) provide sufficiently precise estimates of WIP and LT for cellular lines defined by assumptions (i)-(v) if p i [0.9, 0.99], i =,,..., M, (3.9) p 0 [0.85 min i M p i, 0.99 min i M p i]. (3.0) The analyses presented below are carried out for cellular lines with parameters defined by (3.9) and (3.0). 0

12 4 Analysis and Open-Loop Control of Lead Time in Cellular Lines with Identical Producing Machines In cellular lines with producing machines having identical efficiency, structural and qualitative properties of LT become especially transparent and instructive. Therefore, in the current section we address this case and in the next one provide a generalization to non-identical machines. 4. Structural properties In cellular lines with p i = p, i =,,..., M, expression (3.6) becomes [ C LT = M Cp 0 + pc 0 ( p C 0 ) ] C(p C p C 0 ). (4.) 3, i.e., To characterize its structural properties, we use the relative workload, ρ, introduced in Section ρ = p 0 p, (4.) and the relative lead time defined as lt := LT M, (4.3) i.e., the lead time in units of the smallest possible lead time (which is M cycle times). In terms of these variables, (4.) becomes lt = (C )p Cρ + (pρ)c (p C ρ C ). (4.4) C( ρ C ) Figure 4. illustrates the behavior of lt as a function of ρ [0.85, ) for various p and C, where the curves for C = are shown for comparison purposes. As one can see, all curves in this figure have a knee beyond which lt grows extremely fast. It is of interest to quantify the position of the knee, ρ knee, on the lt(ρ)-curve. To accomplish this, consider the (ρ, lt)-plane, where a unit interval of ρ-axis corresponds to A > units of lt-axis (in Figure 4., A = 0). Introduce the scaling ratio, α =, and recall that the curvature, κ, of a twice differentiable function, f (x), is given by (see A

13 [3]) κ( f (x)) = f xx ( + f x ) 3. (4.5) C = C = C = C = C = C =.5 (ˆρknee, ltknee).5 (ˆρknee, ltknee).5 (ˆρknee, ltknee) lt.5 lt.5 lt.5 κknee(c = ) = 3.46 κknee(c = ) = 44.7 κknee() = 7.55 κknee(c = ) = 9.49 κknee(c = ) = 3.95 κknee() = ρ 0.95 (a) p = 0.9 κknee(c = ) = 3.78 κknee(c = ) = 9.89 κknee() = ρ 0.95 (b) p = ρ 0.95 (c) p = 0.99 Figure 4.: Behavior of lt as a function of ρ for various p and C The curvature of function α lt(ρ) calculated using expression (4.5) is illustrated in Figure 4. for several values of p and C. The unique maximum of these curves leads to the following: 30 0 C = C = C = C = C = C = κ κ 0 κ ρ (a) p = ρ (b) p = ρ (c) p = 0.99 Figure 4.: Curvature of α lt(ρ) as a function of ρ for various p and C Definition 4. The knee, ˆρ knee, of lt on the (ρ, lt)-plane with the scaling ratio α is a point on [0.85, ), at which the curvature of α lt(ρ) reaches its maximum. The pairs (ˆρ knee, lt(ρ knee )) are indicated in Figure 4. by black dots. Thus, releasing raw material with the rate p 0 < pˆρ knee results in lt below the knee. Clearly, larger p 0 results in larger production rate (see (3.7)). Therefore, it would seem desirable to operate the system close to or at the knee. Robustness of such an operation is investigated next.

14 4. Robustness In this subsection, we evaluate robustness of LT with respect to both release rate, p 0, and producing machine efficiency, p. 4.. Robustness with respect to p 0 Assume that a cellular line nominally operates at the knee of lt, i.e., p 0 = pˆρ knee, lt = lt(ˆρ knee ), PR = Cp 0. (4.6) Assume that in reality the release rate is larger than p 0, i.e., p 0,real > p 0. (4.7) This will, obviously, lead to lt > lt(ˆρ knee ). Assume, finally, that the largest tolerable lt (e.g., due to perishable nature of the goods produced) is lt = K lt(ˆρ knee ), (4.8) where K >. In other words, lt must be no larger than K times lt knee := lt(ˆρ knee ). Denote the release rate that results in lt as p 0,real. Then, robustness of lt with respect to p 0 can be quantified as R p0 = p 0,real p 0 p 0 00%. (4.9) We have investigated R p0 numerically solving equations (4.4) and (4.8) for M = 0. The results are shown in Figure 4.3. From this figure we conclude that robustness of lt with respect to p 0 decreases as a function of C: for, this robustness is roughly twice below that for C =. 3

15 8 6 C = C = 8 6 C = C = 8 6 C = C = Rp0 (%) 4 Rp0(%) 4 Rp0 (%) p (a) K = p (b) K = p (c) K = 0 Figure 4.3: Robustness of lt with respect to p 0 as a function of p for various K and C 4.. Robustness with respect to p Assume again that the cellular line nominally operates at ˆρ knee with nominal p and p 0. In reality, however, the producing machines are less efficient than nominally assumed, i.e., p real < p. (4.0) In this case, ρ real = p 0 p real > ˆρ knee. (4.) This again will result in lt(ρ real ) > lt(ˆρ knee ). Assume, as before, that the largest lt that can be tolerated is given by lt(ρ real ) = K lt(ˆρ knee ), (4.) where K >. Denote the producing machine efficiency that results in lt(ρ real ) as p real. Then, robustness of lt with respect to p can be quantified as R p = p real p p 00%. (4.3) The values of R p have been investigated numerically using expressions (4.4) and (4.) for cellular lines defined by assumptions (i)-(v) with M = 0. The results, shown in Figure 4.4, lead to the conclusion similar to the one derived above: robustness of lt with respect to p for is roughly twice lower than for C =. Thus, the results of this subsection indicate that a plausible reason for long lead time in cellular 4

16 8 6 C = C = 8 6 C = C = 8 6 C = C = Rp(%) 4 Rp(%) 4 Rp(%) p (a) K = p (b) K = p (c) K = 0 Figure 4.4: Robustness of lt with respect to p as a function of p for various K and C lines is their low robustness at or close to the knee. 4.3 Optimal release rate as a function of the desired lead time Since the operation at the knee is undesirable, a question arises: At which point on the lt(ρ)- curve should a cellular line operate? This question can be answered by solving the following constrained optimization problem: Find the release rate ˆp 0 such that lt takes the desired value and PR is maximized. A solution of this open-loop control problem is provided by the following: Theorem 4. Consider the cellular line defined by assumptions (i)-(v) with p [0.9, 0.99] and assume that the desired lead time is lt d. Consider the following polynomial equation with respect to ˆp 0 : Then, ˆp C 0 Clt d ˆp C+ 0 + (C ) ˆp C 0 + Clt d p C ˆp 0 (C )p C = 0. (4.4) this equation has at most two solutions on (0, ); if it has no solutions on (0, ), lt d is viewed as infeasible; if it has a single solution in the interval (3.0), this solution is the optimal release rate, ˆp 0 ; if it has both solutions in the interval (3.0), the largest one is the optimal release rate, ˆp 0. Proof: See the Appendix. The behavior of ˆp 0 as a function of lt d is illustrated in Figure 4.5, with black dots indicating ˆp 0 at the knee. From this figure we conclude: 5

17 For lt d < lt knee, the optimal release rate, ˆp 0 (and, therefore, PR), is a rapidly increasing function of lt d. For lt d > lt knee, ˆp 0 (and, therefore, PR) is practically constant (especially in the case of efficient producing machines and large C). Thus, operating beyond the knee is not only unnecessary (since PR is almost constant), but counter-productive as well (since ŴIP grows without bounds) ˆp C = C = lt d (a) p = 0.9 ˆp C = C = lt d (b) p = 0.95 ˆp C = C = lt d (c) p = 0.99 Figure 4.5: Optimal release rate, ˆp 0, as a function of lt d for various p and C 5 Analysis and Open-Loop Control of Lead Time in Cellular Lines with Non-identical Producing Machines 5. Structural properties To illustrate structural behavior of cellular lines with non-identical machines, we use the modified relative workload, ρ max : ρ max = p 0 min p, (5.) i i M where, as before, p 0 is the release rate and p i, i =,,..., M, is the producing machine efficiency. The relative lead time, lt, remains the same as in (4.3), with LT given by (3.6). Then, keeping p i s constant, we plot lt as a function of ρ max. The results for line L of (3.8) are shown in Figure 5. by the solid curves; the results for lines L and L 3 are similar and omitted due to space limitations. Clearly, lt(ρ max ) has a knee similar to that in the identical machine case. Thus, all structural 6

18 properties investigated in Subsection 4. remain the same for the non-identical machine case as well. lt L, C =.5 L, C = (ˆρ knee, lt knee ) ρ max (a) C = lt 3.5 L, C = 3 L, C =.5 (ˆρ knee, lt knee ) ρ max (b) C = lt 3.5 L, 3 L,.5 (ˆρ knee, lt knee ) ρ max (c) Figure 5.: Behavior of lt for line L as a function of ρ max 5. Robustness For non-identical machines, robustness of lt with respect to p 0 and p i can be investigated in the same manner as in Subsection 4.. However, a different approach is possible as well. Specifically, along with a cellular line with non-identical producing machines, consider an auxiliary line with identical producing machines defined by p min = min i M p i. (5.) For this line, results of Subsection 4. are applicable, and its lt(ρ)-characteristic is plotted in Figure 5. by broken lines, with black dots indicating the knee points. As one can see, these curves provide an upper bound of lt(ρ), and, thus, can be used for quantifying robustness of the latter. Based on the above, we conclude that robustness of lt in the case of non-identical machines remains low. 5.3 Optimal release rate as a function of the desired lead time Theorem 5. Consider the cellular line defined by assumptions (i)-(v) with p i [0.9, 0.99], i, and assume that the desired lead time is LT d. Consider the following polynomial equation of 7

19 order (M + )C with respect to ˆp 0 : M M M ˆp C 0 ( ˆpC 0 ) (p C j ˆp C 0 ) C ˆp 0LT d (p C i ˆp C 0 ) + (C )M Then, i= j=, j i i= (5.3) M (p C i ˆp C 0 ) = 0. i= this equation has at most two solutions on (0, ); if it has no solutions on (0, ), LT d is viewed as infeasible; if it has a single solution in the interval (3.0), this solution is the optimal release rate, ˆp 0 ; if it has both solutions in the interval (3.0), the largest one is the optimal release rate, ˆp 0. Proof: See the Appendix. 6 Deterministic Raw Material Release In some cases, random raw material release may be inconvenient for practical implementations. In such situations, results of Sections 4 and 5 can be used to define strategies for deterministic, e.g., hourly, release. To model the hourly release, let the desired lead time be defined in minutes and denoted as LT d. Then the desired lead time in units of cycle time is given by LT d = LT d, (6.) τ where the cycle time, τ, is also in minutes. For example, if LT d = 0 min, LT d = 0 if τ = min and LT d = 00 if τ = 0. min. Given LT d defined by (6.), the corresponding release rate per cycle, ˆp 0, can be calculated using either (4.4) or (5.3). Then, the hourly release, R, is defined as R = HC ˆp 0 (LT d), (6.) 8

20 where x denotes the largest integer not greater than x, and H is the number of cycle times in an hour, i.e., H = 60. Releasing each hour the amount of raw material defined by (6.), leads to the τ following inequalities: LT( ˆp 0 ) < LT(R ) < LT( ˆp 0 ) + H, (6.3) where LT( ˆp 0 ) and LT(R ) are the lead time estimates for per-cycle and per-hour release, respectively. Multiplying these inequalities by τ gives LT d < LT (R ) < LT d (6.4) The tightness of this bound has been evaluated by simulating cellular lines with ten cells under the hourly release for various τ and LT d. For each case, the simulation runs 5, 000, 000 time slots (including, 500, 000 time slots of warm-up period) and for 0 repetitions. The results are shown in Tables 6. and 6. for identical machine case and non-identical machine case, respectively. These results indicate that, if LT d empirical formula: < 60 min, LT (R ) can be evaluated using the following LT (R ) LT d (6.5) Thus, hourly release leads to a relatively insignificant lead time deterioration. 9

21 Table 6.: Lead time, LT (R ), under hourly release for cellular lines with identical machines (a) p = 0.9 C ρ p 0 lt d τ LT d LT (R ) (b) p = 0.95 C ρ p 0 lt d τ LT d LT (R ) (c) p = 0.99 C ρ p 0 lt d τ LT d LT (R )

22 Table 6.: Lead time, LT (R ), under hourly release for cellular lines with non-identical machines (a) L C ρ max p 0 lt d τ LT d LT (R ) (b) L C ρ max p 0 lt d τ LT d LT (R ) (c) L 3 C ρ max p 0 lt d τ LT d LT (R )

23 7 Closed-Loop Control 7. Scenario As stated in Theorems 4. and 5., releasing raw material with rate ˆp 0 ensures the desired lead time for a given p (in the identical machine case) or given p, p,..., p M (in the non-identical machine case). However, in reality, machine efficiency may be different from nominally assumed for ˆp 0 calculation. To model this situation, suppose that in the nominally identical machine case, the real machine efficiency is as follows: p real i = [0.8, 0.95], i =,,..., M, for p = 0.9, [0.85, 0.99], i =,,..., M, for p = 0.95, (7.) [0.9, 0.99], i =,,..., M, for p = For the non-identical machine case, we assume that p real i = [0.8, 0.95], i =,,..., M. (7.) Under these conditions, releasing raw material according to ˆp 0 may result in LT real, which is dramatically different from LT d, especially if it turns out that ˆp 0 > min i M preal i. To avoid this undesirable behavior, closed-loop control of raw material release can be used. In this section, we introduce a simple feedback release control law and investigate its performance for cellular lines with identical and non-identical machines. 7. Control law Although a number of control policies could be considered, we use here a relay-type law given by ˆp 0 p 0 (n + ) =, if WIP(n) C ˆp 0 LT d, 0, otherwise, (7.3) n = 0,,...,

24 where n is the index of time slot, p 0 (n + ) is the release rate in time slot n +, ˆp 0 is the optimal release rate calculated according to Theorems 4. or 5. (using the nominal machine efficiencies), and WIP(n) is the work-in-process in the system at time slot n. In other words, according to (7.3), raw material is released with the nominal rate ˆp 0, if the total work-in-process in the system is not above that defined by the release rate that guarantees LT d ; otherwise no release takes place. Note that the information on WIP(n) can be maintained by counting parts released into and from the system. For the case of hourly release, (7.3) is modified as follows: HC ˆp 0, if WIP(s) C ˆp 0 R(s + ) = LT d, 0, otherwise, (7.4) s = 0,,..., where s is the index of hour, R(s + ) is the number of parts released into the system at the beginning of hour s +, H is the number of time slots in an hour, x, as before, is the largest integer not greater than x, and WIP(s) is the number of parts in the system at the end of hour s. The performance of these laws, for both identical and non-identical machines, has been analyzed by simulations. The results are described below. 7.3 Performance 7.3. Identical machine case We investigated control law (7.3) for cellular lines with 0 producing cells. The values of p and lt d have been selected from the following sets p = {0.9, 0.95, 0.99}, lt d = {, 3, 4, 5, 6, 7, 8, 9, 0}, (7.5) and the actual machine efficiencies have been determined by selecting p i randomly and equiprobably from intervals (7.). For each pair (p, lt d ) from (7.5), 00 production lines have been generated and the relative lead times analyzed by simulation. For each production line, the simulation runs 3

25 50, 000 time slots (including 5, 000 time slots of warm-up period) and for 0 repetitions. For control law (7.3), the average relative lead times and production losses due to feedback release are shown in Figs. 7. and 7., respectively. As one can see, the relative lead time, lt, is close to the desired lead time, lt d, i.e., the feedback enforces the desired lead time performance. The production losses, quantified by PR loss = PR FB PR nofb PR nofb 00% (where PR FB and PR nofb are the production rates of the system with and without feedback control, respectively), is a decreasing function of lt d ; for lt d 6, PR loss 0. As for control law (7.4), the average relative lead times and production losses due to feedback release are shown in Figs. 7.3 and 7.4, from which we observe that the relative lead time may be up to.5 times of the desired lead time for small lt d and close to it for large lt d ; the production loss is also decreasing and goes to zero as lt d increases. 0 C = C = 0 C = C = 0 C = C = lt lt lt ltd ltd ltd (a) p = 0.9 (b) p = 0.95 (c) p = 0.99 Figure 7.: Relative lt of per-cycle release with feedback control (identical machine case) PRloss(%) C = C = PRloss(%) 4 3 C = C = PRloss(%) C = C = ltd (a) p = ltd (b) p = ltd (c) p = 0.99 Figure 7.: Production loss of per-cycle release with feedback control (identical machine case) 7.3. Non-identical machine case We investigated control laws (7.3) and (7.4) using the three nominal lines given in (3.8). The actual machine efficiencies have been determined by selecting p i randomly and equiprobably from 4

26 0 C = C = 0 C = C = 0 C = C = lt lt lt ltd ltd ltd (a) p = 0.9 (b) p = 0.95 (c) p = 0.99 Figure 7.3: Relative lt of per-hour release with feedback control (identical machine case) PRloss(%) C = C = PRloss(%) C = C = PRloss(%) C = C = ltd (a) p = ltd (b) p = ltd (c) p = 0.99 Figure 7.4: Production loss of per-hour release with feedback control (identical machine case) the interval (7.). The simulation procedure has been the same as in Subsection The results obtained are similar to the identical machine case and are omitted due to space constraints. Thus, feedback control laws (7.3) and (7.4) can be used as a means to combat uncertainty in the producing machine efficiency for both identical and non-identical machine cases. 8 Conclusions and Future Work This paper provides a plausible explanation for excessively long lead time often encountered in cellular lines. Specifically, it shows that lead time as a function of the relative load on the system has a knee-type behavior, and operation at or close to the knee (which is desirable from the point of view of the production rate) has low robustness in systems with a large number of machines per cell. The problem is exacerbated for operating points beyond the knee, where the production rate remains practically constant, but the lead time grows without bounds. To avoid this undesirable behavior, the paper offers a means for calculating the optimal release rate that ensures the desired lead time, while maximizing the production rate, and offers a feedback control law to maintain the 5

27 desired lead time. Numerous problems, however, remain open. Some of them are as follows: Extension of the results obtained to system with Bernoulli machines, which are not necessarily asymptotically reliable (i.e., having their efficiency on (0, )). Extension of the results obtained to system with machines obeying the exponential reliability model. Extension of the results obtained to system with non-markovian machines, e.g., having Weibull, gamma, log-normal, etc., reliability models. Investigation of lead time in cellular assembly systems. Investigation of lead time in re-entrant lines, where problems with excessive lead time are notoriously complex. Last but not least, application of the results obtained in practical systems. Solutions of these problems will lead to a relatively complete theory of lead time for cellular production systems. Appendix Proof of Theorem 3.: To prove the first statement of the theorem, we observe that, since the buffer is infinite, the production rate of the system equals the production rate of the release cell. 6

28 Therefore, PR = = C ( ) C i p i 0 i ( p 0) C i C C! i i!(c i)! pi 0 ( p 0) C i i=0 i= = Cp 0 C i= i=0 (C )! (i )!(C i)! pi 0 ( p 0) C i C (C )! = Cp 0 i!(c i)! pi 0 ( p 0) C i = Cp 0. (A.) To prove the second statement, note that, since p = εk, p i = ( εk) i = εki + O(ε ). (A.) Thus, p i = εki + O(ε ) = εk j + O(ε ) εk(i j) (A.3) = p j εk(i j) + O(ε ), 0 j < i. From assumption (iii), it follows that the number of machines, which are up in cell c i, i = 0,, is j with probability ( ) C j p j i ( p i) C j, j = 0,,..., C. Specifically, the probability that C machines are up in cell c i, i = 0,, is ( ) C p C i ( p i ) = Cp C i ( p i ). (A.4) C Based on (A.3), we have p C i = p j i εk i(c j) + O(ε ), 0 j < C, (A.5) which implies Cp C i C = p C i + j=0 ( p j i εk i(c j) + O(ε ) ) (A.6) 7

29 C = j=0 p j i εk C(C ) i + O(ε ). Thus, ( ) C C ( C = j=0 C = ( p i ) p C i ( p i ) p j i εk C(C ) i + O(ε ) j=0 ) ( p i ) ( p j i ( p C(C ) i) εk i ( = p C C(C ) i εk i εk i + O(ε ) = p C i + O(ε ). Similarly, ( ) C C ( C j p C i ( p i ) = O(ε ), ) p j i ( p i) C j = o(ε ), j C 3. ) ) + O(ε ) (A.7) (A.8) In other words, the probability that C machines are up in cell c i, i = 0,, is p C i ; C machines are up with probability p C i + O(ε ); and C and j C 3 machines are up with probability O(ε ) and o(ε ), respectively. Let P i j denote the transition probability of buffer occupancy from j to i. Then, using the above arguments, P C, j = p C 0 + O(ε ), j = 0,,..., C, P C, j = p C 0 + O(ε ), j = 0,,..., C, P ii = p C 0 pc + ( pc 0 )( pc ) + O(ε ), i C, P i,i = ( p C 0 )pc + O(ε ), i C, (A.9) P i+,i = p C 0 ( pc ) + O(ε ), i C, P i j = 0 or O(ε ) or o(ε ), for other i and j. 8

30 Let ˆP i j denote the estimate of P i j defined by ˆP C, j = p C 0, j = 0,,..., C, ˆP C, j = p C 0, j = 0,,..., C, ˆP ii = p C 0 pc + ( pc 0 )( pc ), i C, ˆP i,i = ( p C 0 )pc, i C, (A.0) ˆP i+,i = p C 0 ( pc ), i C, ˆP i j = 0, for other i and j. Based on the perturbation theory [4], ˆP i j P i j = O(ε ), i, j {0,,...}. (A.) Using (A.0) and the conservation law ˆP i = j=0 ˆP i j ˆP j, i, (A.) we obtain the following balance equations: ˆP i = 0, i = 0,,..., C, ˆP C = ( p C 0 ) ˆP C + ( p C 0 )pc ˆP C, (A.3) ˆP C = p C 0 ˆP C + [ p C 0 pc + ( pc 0 )( pc )] ˆP C + ( p C 0 )pc ˆP C+, ˆP i+ = p C 0 ( pc ) ˆP i + [ p C 0 pc + ( pc 0 )( pc )] ˆP i+ + ( p C 0 )pc ˆP i+, i C. The last three rows of the above equations can be re-written as follows: ˆP C = p C 0 p C ( pc 0 ) ˆP C, ˆP i+ = pc 0 ( pc ) p C ( pc 0 ) ˆP i, i C. (A.4) 9

31 Then, To obtain the solution, let α = pc 0 ( pc ) p C ( pc 0 ). ˆP i = αi C+ p C ˆP C, i C. (A.5) (A.6) Taking into account ˆP i = (A.7) i=0 and ˆP i = 0, i = 0,,..., C (see the first row of (A.3)), we obtain ˆP C + α i C+ ˆP C =. (A.8) i=c p C Since p 0 < p (implying that α < ), we have ˆP C = + α p C α = + pc 0 p C pc 0 = pc 0. (A.9) p C 30

32 Therefore, ŴIP = i=c i ˆP i ( ) = (C ) pc 0 + p C ( = (C ) pc 0 p C ( = (C ) pc 0 p C i=c ) + α C+ ) p C i αi C+ p C ( pc 0 p C ( pc 0 ( )[ + α C+ pc 0 Cα C ] p C p C α + αc+ ( α) ( ) = (C ) pc 0 p C p C ) ) iα i i=c ( )[ + pc 0 Cα p C p C α + ( α ) ] α ( ) = (C ) pc 0 p C + pc 0 p C p C p C [C pc 0 ( pc ) p C + ( p C 0 ( pc ) ) ] pc 0 p C pc 0 ( ) = (C ) pc 0 + C pc 0 + pc 0 ( pc ) p C p C p C (pc pc 0 ) = C + pc 0 p C + pc 0 ( pc ) p C (pc pc 0 ) = C + pc 0 ( pc 0 ) p C, pc 0 (A.0) which coincides with (3.) (having p denoted as p). Using the Little s law [5], we obtain LT = ŴIP PR = C Cp 0 + pc 0 ( p C 0 ) C(p C pc 0 ), (A.) and, due to (A.), ŴIP WIP = O(ε ), LT LT = O(ε ). (A.) To prove Theorem 4., we need the following lemma: 3

33 Lemma A. Function lt(ρ) defined by (4.4) is convex. Proof: First, we re-write (4.4) as follows: lt(ρ) = (C )p Cρ + (pρ)c C + (pρ)c (p C ). (A.3) C( ρ C ) If all three terms of (A.3) are convex, then lt(ρ) is also convex (see [6]). Clearly, the first two terms are convex. So, what remains to prove is that f (ρ) := (pρ)c (p C ) C( ρ C ) (A.4) is also convex. For C =, f (ρ) = p ρ, (A.5) which is convex. For C >, let g(ρ) := ρ C, h(x) := pc x(p C ) x C C (A.6) (A.7) and, therefore, f (ρ) = h ( g(ρ) ). (A.8) In the following, we prove that h(x) is convex, i.e., for all 0 < x, x <, h(x ) + h(x ) h ( x + x ). (A.9) 3

34 From (A.7), we have h(x ) + h(x ) h ( x + x ) [ = (p p C x ) + C x C = (p p C ) { x [ x C C x x C C ( x +x ( C x C ) C C ] )[ ( x +x ) C C x +x ( x +x ) C C ] ] (A.30) Let then K = [ x ( x +x + ( C x C ) C C C x C ] )[ ( x +x ) C C p p C ( C )( C x x C C h(x ) + h(x ) h ( x + x ) { = K x C C + x C C ( x + x [( x + x ) ( C C x C + x ] }. )[ ( x +x ) C C ) C C ( x + x ) + x x ) ]} (x x ) C. ], (A.3) (A.3) Since C C C >, x C is convex. Thus, for any x, x > 0, x C C + x C C ( x + x ) C C 0. (A.33) Also, for any x, x > 0, we have ( x + x ) C (x x ) (C ) (A.34) and x C C + x (x x ) (C ). (A.35) 33

35 Multiplying the above two inequalities gives ( x + x ) ( C C x + x C ) (x x ) C 0. (A.36) Based on (A.33) and (A.36) and observing that K in (A.3) is positive, we conclude that (A.9) holds and, thus, h(x) is convex. Since h(x) is also monotonically increasing (see (A.7)) and g(ρ) is convex (see (A.6)), based on the property of the composition of convex functions [6], f (ρ) is also convex, which implies that lt(ρ) is convex. Proof of Theorem 4.: From (4.4), we have lt d = C C ˆp 0 + ˆpC 0 ( ˆp C 0 ) C(p C ˆp C 0 ). (A.37) Multiplying both sides by C ˆp 0 (p C ˆp C 0 ) gives (4.4). Since from Lemma A., lt(ρ) is convex, lt(p 0 ) is also convex for fixed p (0, ). Thus, (4.4) has at most two solutions on (0, ), which proves the first statement of the theorem. If lt d is smaller than the minimum of lt(p 0 ), (4.4) has no solutions on (0, ) and thus, lt d is infeasible, which proves the second statement. If (4.4) has a unique solution on (0, ) and this solution is in the interval (3.0) or (4.4) has two solutions on (0, ) and only one is in the interval (3.0), then this solution is the optimal release rate, ˆp 0, which proves the third statement. If (4.4) has two solutions in the interval of (3.0), then the largest of them is the optimal release rate, ˆp 0, (due to (3.7)), which completes the proof of the theorem. Proof of Theorem 5.: From (3.6), we have LT d = M [ C i= C ˆp 0 + ˆpC 0 ( ˆp C 0 ) ] C(p C i ˆp C 0 ). (A.38) M Multiplying both sides by C ˆp 0 (p C i ˆp C 0 ) gives (5.3). i= Similar to Lemma A., it can be proved that LT(p 0 ) is convex. Based upon this fact, Theorem 5. is proved in the same manner as Theorem

36 References [] N. Viswanadham and Y. Narahari, Performance Modeling of Automated Manufacturing Systems. Englewood Cliffs, N. J.: Prentice Hall, 99. [] R. G. Askin and C. R. Standridge, Modeling and Analysis of Manufacturing Systems. New York: Wiley, 993. [3] J. A. Buzacott and J. G. Shanthikumar, Stochastic Models of Manufacturing Systems. Englewood Cliffs, N. J.: Prentice Hall, 993. [4] H. T. Papadopoulos, C. Heavey, and J. Browne, Queueing Theory in Manufacturing Systems Analysis and Design. London: Chapman & Hall, 993. [5] S. B. Gershwin, Manufacturing Systems Engineering. Englewood Cliffs, N. J.: PTR Prentice Hall, 994. [6] T. Altiok, Performance Analysis of Manufacturing Systems. New York: Springer, 997. [7] J. Li and S. M. Meerkov, Production Systems Engineering. New York: Springer, 009. [8] G. L. Curry and R. M. Feldman, Manufacturing Systems Modeling and Analysis, nd ed. Springer Verlag, 00. [9] E. Ignall and A. Silver, The output of two-stage system with unreliable machines and limited storage, AIIE Transactions, vol. 9, no. 5, pp , 977. [0] Y. A. Phillis, H. D Angelo, and G. C. Saussy, Analysis of series-parallel production networks without buffers, IEEE Transactions on Reliability, vol. 35, no., pp , 986. [] B. Ancelin and A. Semery, Calcul de la productivité d une ligne intégrée de fabrication: CALIF, un logiciel industriel basé sur une nouvelle heuristique, Automatique-Productique Informatique Industrielle, vol., no. 3, pp ,

37 [] T. Iyama and S. Ito, The maximum production rate for an unbalances multiserver flow line system with finite buffer storage, International Journal of Production Research, vol. 5, pp , 987. [3] K. C. Jeong and Y. D. Kim, An approximation method for performance analysis of assembly/disassembly systems with parallel-machine stations, IIE Transactions, vol. 3, no. 4, pp , 999. [4] A. Patchong and D. Willaeys, Modeling and analysis of an unreliable flow line composed of parallel-machine stages, IIE Transactions, vol. 33, no. 7, pp , 00. [5] J. Li, Modeling and analysis of manufacturing systems with parallel lines, IEEE Transactions on Automatic Control, vol. 49, no. 0, pp , 004. [6] C.-B. Yan and Q. Zhao, A unified effective method for aggregating multi-machine stages in production systems, IEEE Transactions on Automatic Control, vol. 58, no. 7, pp , 03. [7] D. Mitra, Stochastic theory of a fluid model of producers and consumers couples by a buffer, Advances in Applied Probability, vol. 0, no. 3, pp , 988. [8] M. H. Burman, New results in flow line analysis, Ph.D. dissertation, Operations Research Center, EECS, MIT, 995. [9] B. Tan, A three-station merge system with unreliable stations and a shared buffer, Mathematical and Computer Modelling, vol. 33, no. 8-9, pp. 0 06, 00. [0] M. Colledani and S. B. Gershwin, Modeling and analysis of two-stage systems with parallel machines and limited repair capacity, in Preprints of the 3th IFAC Symposium on Information Control Problems in Manufacturing, Moscow, Russia, 009, pp [] B. Tan and S. B. Gershwin, Analysis of a general markovian two-stage continuous-flow production system with a finite buffer, International Journal of Production Economics, vol. 0, no., pp ,

38 [] S. Biller, S. M. Meerkov, and C.-B. Yan, Raw material release rates to ensure desired production lead time in Bernoulli serial lines, International Journal of Production Research, vol. 5, no. 4, pp , 03. [3] R. Courant and F. John, Introduction to calculus and analysis. Springer, 999, vol.. [4] R. Bellman, Introduction to Matrix Analysis, nd ed. SIAM, 997. [5] J. D. C. Little, A proof for the queuing formula: L = λw, Operations Research, vol. 9, no. 3, pp , 96. [6] S. Boyd and L. Vandenberghe, Convex Optimization. Cambridge University Press,

A Method for Sweet Point Operation of Re-entrant Lines

A Method for Sweet Point Operation of Re-entrant Lines A Method for Sweet Point Operation of Re-entrant Lines Semyon M. Meerkov, Chao-Bo Yan, Department of Electrical Engineering and Computer Science, University of Michigan, Ann Arbor, MI 4819-2122, USA (e-mail:

More information

PERFORMANCE ANALYSIS OF PRODUCTION SYSTEMS WITH REWORK LOOPS

PERFORMANCE ANALYSIS OF PRODUCTION SYSTEMS WITH REWORK LOOPS PERFORMANCE ANALYSIS OF PRODUCTION SYSTEMS WITH REWORK LOOPS Jingshan Li Enterprise Systems Laboratory General Motors Research & Development Center Mail Code 480-106-359 30500 Mound Road Warren, MI 48090-9055

More information

MODELING AND ANALYSIS OF SPLIT AND MERGE PRODUCTION SYSTEMS

MODELING AND ANALYSIS OF SPLIT AND MERGE PRODUCTION SYSTEMS University of Kentucky UKnowledge University of Kentucky Master's Theses Graduate School 008 MODELING AND ANALYSIS OF SPLIT AND MERGE PRODUCTION SYSTEMS Yang Liu University of Kentucky, yang.liu@uky.edu

More information

Department 2 Power cut of A1 and A2

Department 2 Power cut of A1 and A2 3.12. PROBLEMS 115 3.12 Problems Problem 3.1 A production system manufactures products A and B. Each product consists of two parts: A1 and A2 for product A and B1 and B2 for product B. The processing of

More information

Production variability in manufacturing systems: Bernoulli reliability case

Production variability in manufacturing systems: Bernoulli reliability case Annals of Operations Research 93 (2000) 299 324 299 Production variability in manufacturing systems: Bernoulli reliability case Jingshan Li and Semyon M. Meerkov Department of Electrical Engineering and

More information

Bottlenecks in Markovian Production Lines: A Systems Approach

Bottlenecks in Markovian Production Lines: A Systems Approach 352 IEEE TRANSACTIONS ON ROBOTICS AND AUTOMATION, VOL. 14, NO. 2, APRIL 1998 Bottlenecks in Markovian Production Lines: A Systems Approach S.-Y. Chiang, C.-T. Kuo, and S. M. Meerkov Abstract In this paper,

More information

MIT Manufacturing Systems Analysis Lectures 6 9: Flow Lines

MIT Manufacturing Systems Analysis Lectures 6 9: Flow Lines 2.852 Manufacturing Systems Analysis 1/165 Copyright 2010 c Stanley B. Gershwin. MIT 2.852 Manufacturing Systems Analysis Lectures 6 9: Flow Lines Models That Can Be Analyzed Exactly Stanley B. Gershwin

More information

SCHEDULING POLICIES IN MULTI-PRODUCT MANUFACTURING SYSTEMS WITH SEQUENCE-DEPENDENT SETUP TIMES

SCHEDULING POLICIES IN MULTI-PRODUCT MANUFACTURING SYSTEMS WITH SEQUENCE-DEPENDENT SETUP TIMES Proceedings of the 2011 Winter Simulation Conference S. Jain, R. R. Creasey, J. Himmelspach, K. P. White, and M. Fu, eds. SCHEDULING POLICIES IN MULTI-PRODUCT MANUFACTURING SYSTEMS WITH SEQUENCE-DEPENDENT

More information

Bu er capacity for accommodating machine downtime in serial production lines

Bu er capacity for accommodating machine downtime in serial production lines int. j. prod. res., 2002, vol. 40, no. 3, 601±624 Bu er capacity for accommodating machine downtime in serial production lines EMRE ENGINARLARy, JINGSHAN LIz, SEMYON M. MEERKOVy* and RACHEL Q. ZHANG This

More information

A Bernoulli Model of Selective Assembly Systems

A Bernoulli Model of Selective Assembly Systems Preprints of the 19th World Congress The International Federation of Automatic Control A Bernoulli Model of Selective Assembly Systems Feng Ju and Jingshan Li Department of Industrial and Systems Engineering,

More information

Glossary availability cellular manufacturing closed queueing network coefficient of variation (CV) conditional probability CONWIP

Glossary availability cellular manufacturing closed queueing network coefficient of variation (CV) conditional probability CONWIP Glossary availability The long-run average fraction of time that the processor is available for processing jobs, denoted by a (p. 113). cellular manufacturing The concept of organizing the factory into

More information

Markovian Modeling of Multiclass Deterministic Flow Lines with Random Arrivals: The Case of a Single-Channel

Markovian Modeling of Multiclass Deterministic Flow Lines with Random Arrivals: The Case of a Single-Channel 2015 IEEE International Conference on Automation Science and Engineering (CASE) Aug 24-28, 2015. Gothenburg, Sweden Markovian Modeling of Multiclass Deterministic Flow Lines with Random Arrivals: The Case

More information

ACHIEVING OPTIMAL DESIGN OF THE PRODUCTION LINE WITH OBTAINABLE RESOURCE CAPACITY. Miao-Sheng CHEN. Chun-Hsiung LAN

ACHIEVING OPTIMAL DESIGN OF THE PRODUCTION LINE WITH OBTAINABLE RESOURCE CAPACITY. Miao-Sheng CHEN. Chun-Hsiung LAN Yugoslav Journal of Operations Research 12 (2002), Number 2, 203-214 ACHIEVING OPTIMAL DESIGN OF THE PRODUCTION LINE WITH OBTAINABLE RESOURCE CAPACITY Miao-Sheng CHEN Graduate Institute of Management Nanhua

More information

Representation and Analysis of Transfer Lines. with Machines That Have Different Processing Rates

Representation and Analysis of Transfer Lines. with Machines That Have Different Processing Rates March, 1985 Revised November, 1985 LIDS-P-1446 Representation and Analysis of Transfer Lines with Machines That Have Different Processing Rates by Stanley B. Gershwin 35-433 Massachusetts Institute of

More information

Transient Analysis of Single Machine Production Line Dynamics

Transient Analysis of Single Machine Production Line Dynamics International Journal of Operations Research International Journal of Operations Research Vol. 11, No. 2, 040 050 (2014) Transient Analysis of Single Machine Production Line Dynamics Farhood Rismanchian

More information

2 Theory. 2.1 State Space Representation S 2 S 1 S 3

2 Theory. 2.1 State Space Representation S 2 S 1 S 3 In the following sections we develop the theory, illustrate the technique by applying it to a sample system, and validate the results using the method of enumeration. Notations: A-state functional (acceptable)

More information

An Aggregation Method for Performance Evaluation of a Tandem Homogenous Production Line with Machines Having Multiple Failure Modes

An Aggregation Method for Performance Evaluation of a Tandem Homogenous Production Line with Machines Having Multiple Failure Modes An Aggregation Method for Performance Evaluation of a Tandem Homogenous Production Line with Machines Having Multiple Failure Modes Ahmed-Tidjani Belmansour Mustapha Nourelfath November 2008 CIRRELT-2008-53

More information

A Starvation-free Algorithm For Achieving 100% Throughput in an Input- Queued Switch

A Starvation-free Algorithm For Achieving 100% Throughput in an Input- Queued Switch A Starvation-free Algorithm For Achieving 00% Throughput in an Input- Queued Switch Abstract Adisak ekkittikul ick ckeown Department of Electrical Engineering Stanford University Stanford CA 9405-400 Tel

More information

ON THE COEFFICIENTS OF VARIATION OF UPTIME AND DOWNTIME IN MANUFACTURING EQUIPMENT

ON THE COEFFICIENTS OF VARIATION OF UPTIME AND DOWNTIME IN MANUFACTURING EQUIPMENT ON THE COEFFICIENTS OF VARIATION OF UPTIME AND DOWNTIME IN MANUFACTURING EQUIPMENT JINGSHAN LI AND SEMYON M. MEERKOV Received 28 November 24 It was reported in the literature that the coefficients of variation

More information

Single-part-type, multiple stage systems. Lecturer: Stanley B. Gershwin

Single-part-type, multiple stage systems. Lecturer: Stanley B. Gershwin Single-part-type, multiple stage systems Lecturer: Stanley B. Gershwin Flow Line... also known as a Production or Transfer Line. M 1 B 1 M 2 B 2 M 3 B 3 M 4 B 4 M 5 B 5 M 6 Machine Buffer Machines are

More information

Prediction-based adaptive control of a class of discrete-time nonlinear systems with nonlinear growth rate

Prediction-based adaptive control of a class of discrete-time nonlinear systems with nonlinear growth rate www.scichina.com info.scichina.com www.springerlin.com Prediction-based adaptive control of a class of discrete-time nonlinear systems with nonlinear growth rate WEI Chen & CHEN ZongJi School of Automation

More information

Production Capacity Modeling of Alternative, Nonidentical, Flexible Machines

Production Capacity Modeling of Alternative, Nonidentical, Flexible Machines The International Journal of Flexible Manufacturing Systems, 14, 345 359, 2002 c 2002 Kluwer Academic Publishers Manufactured in The Netherlands Production Capacity Modeling of Alternative, Nonidentical,

More information

Optimal Polynomial Control for Discrete-Time Systems

Optimal Polynomial Control for Discrete-Time Systems 1 Optimal Polynomial Control for Discrete-Time Systems Prof Guy Beale Electrical and Computer Engineering Department George Mason University Fairfax, Virginia Correspondence concerning this paper should

More information

Efficient Nonlinear Optimizations of Queuing Systems

Efficient Nonlinear Optimizations of Queuing Systems Efficient Nonlinear Optimizations of Queuing Systems Mung Chiang, Arak Sutivong, and Stephen Boyd Electrical Engineering Department, Stanford University, CA 9435 Abstract We present a systematic treatment

More information

Proceedings of the 2016 Winter Simulation Conference T. M. K. Roeder, P. I. Frazier, R. Szechtman, E. Zhou, T. Huschka, and S. E. Chick, eds.

Proceedings of the 2016 Winter Simulation Conference T. M. K. Roeder, P. I. Frazier, R. Szechtman, E. Zhou, T. Huschka, and S. E. Chick, eds. Proceedings of the 2016 Winter Simulation Conference T. M. K. Roeder, P. I. Frazier, R. Szechtman, E. Zhou, T. Huscha, and S. E. Chic, eds. OPTIMAL COMPUTING BUDGET ALLOCATION WITH EXPONENTIAL UNDERLYING

More information

Transient Performance Evaluation, Bottleneck Analysis and Control of Production Systems

Transient Performance Evaluation, Bottleneck Analysis and Control of Production Systems University of Connecticut DigitalCommons@UConn Doctoral Dissertations University of Connecticut Graduate School 4-19-2017 Transient Performance Evaluation, Bottleneck Analysis and Control of Production

More information

Operations Research Letters. Instability of FIFO in a simple queueing system with arbitrarily low loads

Operations Research Letters. Instability of FIFO in a simple queueing system with arbitrarily low loads Operations Research Letters 37 (2009) 312 316 Contents lists available at ScienceDirect Operations Research Letters journal homepage: www.elsevier.com/locate/orl Instability of FIFO in a simple queueing

More information

Robust Stabilization of Non-Minimum Phase Nonlinear Systems Using Extended High Gain Observers

Robust Stabilization of Non-Minimum Phase Nonlinear Systems Using Extended High Gain Observers 28 American Control Conference Westin Seattle Hotel, Seattle, Washington, USA June 11-13, 28 WeC15.1 Robust Stabilization of Non-Minimum Phase Nonlinear Systems Using Extended High Gain Observers Shahid

More information

Improved Algorithms for Machine Allocation in Manufacturing Systems

Improved Algorithms for Machine Allocation in Manufacturing Systems Improved Algorithms for Machine Allocation in Manufacturing Systems Hans Frenk Martine Labbé Mario van Vliet Shuzhong Zhang October, 1992 Econometric Institute, Erasmus University Rotterdam, the Netherlands.

More information

Unsupervised Learning with Permuted Data

Unsupervised Learning with Permuted Data Unsupervised Learning with Permuted Data Sergey Kirshner skirshne@ics.uci.edu Sridevi Parise sparise@ics.uci.edu Padhraic Smyth smyth@ics.uci.edu School of Information and Computer Science, University

More information

UNIVERSITY OF CALGARY. A Method for Stationary Analysis and Control during Transience in Multi-State Stochastic. Manufacturing Systems

UNIVERSITY OF CALGARY. A Method for Stationary Analysis and Control during Transience in Multi-State Stochastic. Manufacturing Systems UNIVERSITY OF CALGARY A Method for Stationary Analysis and Control during Transience in Multi-State Stochastic Manufacturing Systems by Alireza Fazlirad A THESIS SUBMITTED TO THE FACULTY OF GRADUATE STUDIES

More information

0utline. 1. Tools from Operations Research. 2. Applications

0utline. 1. Tools from Operations Research. 2. Applications 0utline 1. Tools from Operations Research Little s Law (average values) Unreliable Machine(s) (operation dependent) Buffers (zero buffers & infinite buffers) M/M/1 Queue (effects of variation) 2. Applications

More information

FINITE HORIZON ROBUST MODEL PREDICTIVE CONTROL USING LINEAR MATRIX INEQUALITIES. Danlei Chu, Tongwen Chen, Horacio J. Marquez

FINITE HORIZON ROBUST MODEL PREDICTIVE CONTROL USING LINEAR MATRIX INEQUALITIES. Danlei Chu, Tongwen Chen, Horacio J. Marquez FINITE HORIZON ROBUST MODEL PREDICTIVE CONTROL USING LINEAR MATRIX INEQUALITIES Danlei Chu Tongwen Chen Horacio J Marquez Department of Electrical and Computer Engineering University of Alberta Edmonton

More information

PRODUCTVIVITY IMPROVEMENT MODEL OF AN UNRELIABLE THREE STATIONS AND A SINGLE BUFFER WITH CONTINUOUS MATERIAL FLOW

PRODUCTVIVITY IMPROVEMENT MODEL OF AN UNRELIABLE THREE STATIONS AND A SINGLE BUFFER WITH CONTINUOUS MATERIAL FLOW Vol. 3, No. 1, Fall, 2016, pp. 23-40 ISSN 2158-835X (print), 2158-8368 (online), All Rights Reserved PRODUCTVIVITY IMPROVEMENT MODEL OF AN UNRELIABLE THREE STATIONS AND A SINGLE BUFFER WITH CONTINUOUS

More information

MODELING AND PROPERTIES OF GENERALIZED KANBAN CONTROLLED ASSEMBLY SYSTEMS 1

MODELING AND PROPERTIES OF GENERALIZED KANBAN CONTROLLED ASSEMBLY SYSTEMS 1 MODELING AND PROPERTIES OF GENERALIZED KANBAN CONTROLLED ASSEMBLY SYSTEMS 1 Sbiti N. *, Di Mascolo M. **, Amghar M.* * Ecole Mohammadia d Ingénieurs, Université Mohammed V-Agdal, Avenue Ibn Sina, BP 765,

More information

Exponential Disutility Functions in Transportation Problems: A New Theoretical Justification

Exponential Disutility Functions in Transportation Problems: A New Theoretical Justification University of Texas at El Paso DigitalCommons@UTEP Departmental Technical Reports (CS) Department of Computer Science 1-1-2007 Exponential Disutility Functions in Transportation Problems: A New Theoretical

More information

Design of Plant Layouts with Queueing Effects

Design of Plant Layouts with Queueing Effects Design of Plant Layouts with Queueing Effects Saifallah Benjaafar Department of echanical Engineering University of innesota inneapolis, N 55455 July 10, 1997 Abstract In this paper, we present a formulation

More information

On the application of different numerical methods to obtain null-spaces of polynomial matrices. Part 1: block Toeplitz algorithms.

On the application of different numerical methods to obtain null-spaces of polynomial matrices. Part 1: block Toeplitz algorithms. On the application of different numerical methods to obtain null-spaces of polynomial matrices. Part 1: block Toeplitz algorithms. J.C. Zúñiga and D. Henrion Abstract Four different algorithms are designed

More information

THE BOUNDING DISCRETE PHASE TYPE METHOD

THE BOUNDING DISCRETE PHASE TYPE METHOD THE BOUNDING DISCRETE PHASE TYPE METHOD JEAN SÉBASTIEN TANCREZ AND PIERRE SEMAL Abstract. Models of production systems have always been essential. They are needed at a strategic level in order to guide

More information

Error Exponent Region for Gaussian Broadcast Channels

Error Exponent Region for Gaussian Broadcast Channels Error Exponent Region for Gaussian Broadcast Channels Lihua Weng, S. Sandeep Pradhan, and Achilleas Anastasopoulos Electrical Engineering and Computer Science Dept. University of Michigan, Ann Arbor, MI

More information

A Nonlinear Programming Approach For a Fuzzy queue with an unreliable server Dr.V. Ashok Kumar

A Nonlinear Programming Approach For a Fuzzy queue with an unreliable server Dr.V. Ashok Kumar The Bulletin of Society for Mathematical Services and Standards Online: 2012-06-04 ISSN: 2277-8020, Vol. 2, pp 44-56 doi:10.18052/www.scipress.com/bsmass.2.44 2012 SciPress Ltd., Switzerland A Nonlinear

More information

On the Throughput, Capacity and Stability Regions of Random Multiple Access over Standard Multi-Packet Reception Channels

On the Throughput, Capacity and Stability Regions of Random Multiple Access over Standard Multi-Packet Reception Channels On the Throughput, Capacity and Stability Regions of Random Multiple Access over Standard Multi-Packet Reception Channels Jie Luo, Anthony Ephremides ECE Dept. Univ. of Maryland College Park, MD 20742

More information

Chapter 2 Optimal Control Problem

Chapter 2 Optimal Control Problem Chapter 2 Optimal Control Problem Optimal control of any process can be achieved either in open or closed loop. In the following two chapters we concentrate mainly on the first class. The first chapter

More information

Selected Examples of CONIC DUALITY AT WORK Robust Linear Optimization Synthesis of Linear Controllers Matrix Cube Theorem A.

Selected Examples of CONIC DUALITY AT WORK Robust Linear Optimization Synthesis of Linear Controllers Matrix Cube Theorem A. . Selected Examples of CONIC DUALITY AT WORK Robust Linear Optimization Synthesis of Linear Controllers Matrix Cube Theorem A. Nemirovski Arkadi.Nemirovski@isye.gatech.edu Linear Optimization Problem,

More information

On the convergence of normal forms for analytic control systems

On the convergence of normal forms for analytic control systems Chapter 1 On the convergence of normal forms for analytic control systems Wei Kang Department of Mathematics, Naval Postgraduate School Monterey, CA 93943 wkang@nps.navy.mil Arthur J. Krener Department

More information

Structural Reliability

Structural Reliability Structural Reliability Thuong Van DANG May 28, 2018 1 / 41 2 / 41 Introduction to Structural Reliability Concept of Limit State and Reliability Review of Probability Theory First Order Second Moment Method

More information

Queueing Theory I Summary! Little s Law! Queueing System Notation! Stationary Analysis of Elementary Queueing Systems " M/M/1 " M/M/m " M/M/1/K "

Queueing Theory I Summary! Little s Law! Queueing System Notation! Stationary Analysis of Elementary Queueing Systems  M/M/1  M/M/m  M/M/1/K Queueing Theory I Summary Little s Law Queueing System Notation Stationary Analysis of Elementary Queueing Systems " M/M/1 " M/M/m " M/M/1/K " Little s Law a(t): the process that counts the number of arrivals

More information

Multi-channel Opportunistic Access: A Case of Restless Bandits with Multiple Plays

Multi-channel Opportunistic Access: A Case of Restless Bandits with Multiple Plays Multi-channel Opportunistic Access: A Case of Restless Bandits with Multiple Plays Sahand Haji Ali Ahmad, Mingyan Liu Abstract This paper considers the following stochastic control problem that arises

More information

Adaptive Dual Control

Adaptive Dual Control Adaptive Dual Control Björn Wittenmark Department of Automatic Control, Lund Institute of Technology Box 118, S-221 00 Lund, Sweden email: bjorn@control.lth.se Keywords: Dual control, stochastic control,

More information

SOLVING POWER AND OPTIMAL POWER FLOW PROBLEMS IN THE PRESENCE OF UNCERTAINTY BY AFFINE ARITHMETIC

SOLVING POWER AND OPTIMAL POWER FLOW PROBLEMS IN THE PRESENCE OF UNCERTAINTY BY AFFINE ARITHMETIC SOLVING POWER AND OPTIMAL POWER FLOW PROBLEMS IN THE PRESENCE OF UNCERTAINTY BY AFFINE ARITHMETIC Alfredo Vaccaro RTSI 2015 - September 16-18, 2015, Torino, Italy RESEARCH MOTIVATIONS Power Flow (PF) and

More information

Proceedings of the 2016 Winter Simulation Conference T. M. K. Roeder, P. I. Frazier, R. Szechtman, E. Zhou, T. Huschka, and S. E. Chick, eds.

Proceedings of the 2016 Winter Simulation Conference T. M. K. Roeder, P. I. Frazier, R. Szechtman, E. Zhou, T. Huschka, and S. E. Chick, eds. Proceedings of the 2016 Winter Simulation Conference T. M. K. Roeder, P. I. Frazier, R. Szechtman, E. Zhou, T. Huschka, and S. E. Chick, eds. EXTENDED KERNEL REGRESSION: A MULTI-RESOLUTION METHOD TO COMBINE

More information

Small Gain Theorems on Input-to-Output Stability

Small Gain Theorems on Input-to-Output Stability Small Gain Theorems on Input-to-Output Stability Zhong-Ping Jiang Yuan Wang. Dept. of Electrical & Computer Engineering Polytechnic University Brooklyn, NY 11201, U.S.A. zjiang@control.poly.edu Dept. of

More information

Slides 9: Queuing Models

Slides 9: Queuing Models Slides 9: Queuing Models Purpose Simulation is often used in the analysis of queuing models. A simple but typical queuing model is: Queuing models provide the analyst with a powerful tool for designing

More information

IEEE TRANSACTIONS ON AUTOMATIC CONTROL, VOL. 43, NO. 3, MARCH

IEEE TRANSACTIONS ON AUTOMATIC CONTROL, VOL. 43, NO. 3, MARCH IEEE TRANSACTIONS ON AUTOMATIC CONTROL, VOL. 43, NO. 3, MARCH 1998 315 Asymptotic Buffer Overflow Probabilities in Multiclass Multiplexers: An Optimal Control Approach Dimitris Bertsimas, Ioannis Ch. Paschalidis,

More information

I. D. Landau, A. Karimi: A Course on Adaptive Control Adaptive Control. Part 9: Adaptive Control with Multiple Models and Switching

I. D. Landau, A. Karimi: A Course on Adaptive Control Adaptive Control. Part 9: Adaptive Control with Multiple Models and Switching I. D. Landau, A. Karimi: A Course on Adaptive Control - 5 1 Adaptive Control Part 9: Adaptive Control with Multiple Models and Switching I. D. Landau, A. Karimi: A Course on Adaptive Control - 5 2 Outline

More information

EECE Adaptive Control

EECE Adaptive Control EECE 574 - Adaptive Control Overview Guy Dumont Department of Electrical and Computer Engineering University of British Columbia Lectures: Thursday 09h00-12h00 Location: PPC 101 Guy Dumont (UBC) EECE 574

More information

On the errors introduced by the naive Bayes independence assumption

On the errors introduced by the naive Bayes independence assumption On the errors introduced by the naive Bayes independence assumption Author Matthijs de Wachter 3671100 Utrecht University Master Thesis Artificial Intelligence Supervisor Dr. Silja Renooij Department of

More information

Stochastic Modeling and Analysis of Generalized Kanban Controlled Unsaturated finite capacitated Multi-Stage Production System

Stochastic Modeling and Analysis of Generalized Kanban Controlled Unsaturated finite capacitated Multi-Stage Production System Stochastic Modeling and Analysis of Generalized anban Controlled Unsaturated finite capacitated Multi-Stage Production System Mitnala Sreenivasa Rao 1 orada Viswanatha Sharma 2 1 Department of Mechanical

More information

THIS paper deals with robust control in the setup associated

THIS paper deals with robust control in the setup associated IEEE TRANSACTIONS ON AUTOMATIC CONTROL, VOL 50, NO 10, OCTOBER 2005 1501 Control-Oriented Model Validation and Errors Quantification in the `1 Setup V F Sokolov Abstract A priori information required for

More information

Quantifying Stochastic Model Errors via Robust Optimization

Quantifying Stochastic Model Errors via Robust Optimization Quantifying Stochastic Model Errors via Robust Optimization IPAM Workshop on Uncertainty Quantification for Multiscale Stochastic Systems and Applications Jan 19, 2016 Henry Lam Industrial & Operations

More information

Bounds on Achievable Rates of LDPC Codes Used Over the Binary Erasure Channel

Bounds on Achievable Rates of LDPC Codes Used Over the Binary Erasure Channel Bounds on Achievable Rates of LDPC Codes Used Over the Binary Erasure Channel Ohad Barak, David Burshtein and Meir Feder School of Electrical Engineering Tel-Aviv University Tel-Aviv 69978, Israel Abstract

More information

Randomized Algorithms

Randomized Algorithms Randomized Algorithms Prof. Tapio Elomaa tapio.elomaa@tut.fi Course Basics A new 4 credit unit course Part of Theoretical Computer Science courses at the Department of Mathematics There will be 4 hours

More information

The ϵ-capacity of a gain matrix and tolerable disturbances: Discrete-time perturbed linear systems

The ϵ-capacity of a gain matrix and tolerable disturbances: Discrete-time perturbed linear systems IOSR Journal of Mathematics (IOSR-JM) e-issn: 2278-5728, p-issn: 2319-765X. Volume 11, Issue 3 Ver. IV (May - Jun. 2015), PP 52-62 www.iosrjournals.org The ϵ-capacity of a gain matrix and tolerable disturbances:

More information

On the Partitioning of Servers in Queueing Systems during Rush Hour

On the Partitioning of Servers in Queueing Systems during Rush Hour On the Partitioning of Servers in Queueing Systems during Rush Hour Bin Hu Saif Benjaafar Department of Operations and Management Science, Ross School of Business, University of Michigan at Ann Arbor,

More information

IMPROVED MPC DESIGN BASED ON SATURATING CONTROL LAWS

IMPROVED MPC DESIGN BASED ON SATURATING CONTROL LAWS IMPROVED MPC DESIGN BASED ON SATURATING CONTROL LAWS D. Limon, J.M. Gomes da Silva Jr., T. Alamo and E.F. Camacho Dpto. de Ingenieria de Sistemas y Automática. Universidad de Sevilla Camino de los Descubrimientos

More information

Upper Bounds on the Capacity of Binary Intermittent Communication

Upper Bounds on the Capacity of Binary Intermittent Communication Upper Bounds on the Capacity of Binary Intermittent Communication Mostafa Khoshnevisan and J. Nicholas Laneman Department of Electrical Engineering University of Notre Dame Notre Dame, Indiana 46556 Email:{mhoshne,

More information

A Novel Integral-Based Event Triggering Control for Linear Time-Invariant Systems

A Novel Integral-Based Event Triggering Control for Linear Time-Invariant Systems 53rd IEEE Conference on Decision and Control December 15-17, 2014. Los Angeles, California, USA A Novel Integral-Based Event Triggering Control for Linear Time-Invariant Systems Seyed Hossein Mousavi 1,

More information

An engineering approximation for the mean waiting time in the M/H 2 b /s queue

An engineering approximation for the mean waiting time in the M/H 2 b /s queue An engineering approximation for the mean waiting time in the M/H b /s queue Francisco Barceló Universidad Politécnica de Catalunya c/ Jordi Girona, -3, Barcelona 08034 Email : barcelo@entel.upc.es Abstract

More information

Combining Shared Coin Algorithms

Combining Shared Coin Algorithms Combining Shared Coin Algorithms James Aspnes Hagit Attiya Keren Censor Abstract This paper shows that shared coin algorithms can be combined to optimize several complexity measures, even in the presence

More information

One important issue in the study of queueing systems is to characterize departure processes. Study on departure processes was rst initiated by Burke (

One important issue in the study of queueing systems is to characterize departure processes. Study on departure processes was rst initiated by Burke ( The Departure Process of the GI/G/ Queue and Its MacLaurin Series Jian-Qiang Hu Department of Manufacturing Engineering Boston University 5 St. Mary's Street Brookline, MA 2446 Email: hqiang@bu.edu June

More information

Synchronized Queues with Deterministic Arrivals

Synchronized Queues with Deterministic Arrivals Synchronized Queues with Deterministic Arrivals Dimitra Pinotsi and Michael A. Zazanis Department of Statistics Athens University of Economics and Business 76 Patission str., Athens 14 34, Greece Abstract

More information

Optimal State Estimation for Boolean Dynamical Systems using a Boolean Kalman Smoother

Optimal State Estimation for Boolean Dynamical Systems using a Boolean Kalman Smoother Optimal State Estimation for Boolean Dynamical Systems using a Boolean Kalman Smoother Mahdi Imani and Ulisses Braga-Neto Department of Electrical and Computer Engineering Texas A&M University College

More information

Machine Learning 4771

Machine Learning 4771 Machine Learning 477 Instructor: Tony Jebara Topic 5 Generalization Guarantees VC-Dimension Nearest Neighbor Classification (infinite VC dimension) Structural Risk Minimization Support Vector Machines

More information

Optimization Tutorial 1. Basic Gradient Descent

Optimization Tutorial 1. Basic Gradient Descent E0 270 Machine Learning Jan 16, 2015 Optimization Tutorial 1 Basic Gradient Descent Lecture by Harikrishna Narasimhan Note: This tutorial shall assume background in elementary calculus and linear algebra.

More information

Available online at ScienceDirect. Procedia Engineering 100 (2015 )

Available online at   ScienceDirect. Procedia Engineering 100 (2015 ) Available online at www.sciencedirect.com ScienceDirect Procedia Engineering 100 (015 ) 345 349 5th DAAAM International Symposium on Intelligent Manufacturing and Automation, DAAAM 014 Control of Airflow

More information

Introduction to Techniques for Counting

Introduction to Techniques for Counting Introduction to Techniques for Counting A generating function is a device somewhat similar to a bag. Instead of carrying many little objects detachedly, which could be embarrassing, we put them all in

More information

LEAN BUFFERING IN SERIAL PRODUCTION LINES WITH BERNOULLI MACHINES

LEAN BUFFERING IN SERIAL PRODUCTION LINES WITH BERNOULLI MACHINES LEAN BUFFERING IN SERIAL PRODUCTION LINES WITH BERNOULLI MACHINES A. B. HU AND S. M. MEERKOV Received 8 February 26; Accepted 7 April 26 Lean buffering is the smallest buffer capacity necessary to ensure

More information

Prioritized Sweeping Converges to the Optimal Value Function

Prioritized Sweeping Converges to the Optimal Value Function Technical Report DCS-TR-631 Prioritized Sweeping Converges to the Optimal Value Function Lihong Li and Michael L. Littman {lihong,mlittman}@cs.rutgers.edu RL 3 Laboratory Department of Computer Science

More information

Oblivious Equilibrium: A Mean Field Approximation for Large-Scale Dynamic Games

Oblivious Equilibrium: A Mean Field Approximation for Large-Scale Dynamic Games Oblivious Equilibrium: A Mean Field Approximation for Large-Scale Dynamic Games Gabriel Y. Weintraub, Lanier Benkard, and Benjamin Van Roy Stanford University {gweintra,lanierb,bvr}@stanford.edu Abstract

More information

Hegselmann-Krause Dynamics: An Upper Bound on Termination Time

Hegselmann-Krause Dynamics: An Upper Bound on Termination Time Hegselmann-Krause Dynamics: An Upper Bound on Termination Time B. Touri Coordinated Science Laboratory University of Illinois Urbana, IL 680 touri@illinois.edu A. Nedić Industrial and Enterprise Systems

More information

Diagnosis of Repeated/Intermittent Failures in Discrete Event Systems

Diagnosis of Repeated/Intermittent Failures in Discrete Event Systems Diagnosis of Repeated/Intermittent Failures in Discrete Event Systems Shengbing Jiang, Ratnesh Kumar, and Humberto E. Garcia Abstract We introduce the notion of repeated failure diagnosability for diagnosing

More information

Singular perturbation analysis of an additive increase multiplicative decrease control algorithm under time-varying buffering delays.

Singular perturbation analysis of an additive increase multiplicative decrease control algorithm under time-varying buffering delays. Singular perturbation analysis of an additive increase multiplicative decrease control algorithm under time-varying buffering delays. V. Guffens 1 and G. Bastin 2 Intelligent Systems and Networks Research

More information

A POMDP Framework for Cognitive MAC Based on Primary Feedback Exploitation

A POMDP Framework for Cognitive MAC Based on Primary Feedback Exploitation A POMDP Framework for Cognitive MAC Based on Primary Feedback Exploitation Karim G. Seddik and Amr A. El-Sherif 2 Electronics and Communications Engineering Department, American University in Cairo, New

More information

Design and Stability Analysis of Single-Input Fuzzy Logic Controller

Design and Stability Analysis of Single-Input Fuzzy Logic Controller IEEE TRANSACTIONS ON SYSTEMS, MAN, AND CYBERNETICS PART B: CYBERNETICS, VOL. 30, NO. 2, APRIL 2000 303 Design and Stability Analysis of Single-Input Fuzzy Logic Controller Byung-Jae Choi, Seong-Woo Kwak,

More information

A PARAMETRIC DECOMPOSITION BASED APPROACH FOR MULTI-CLASS CLOSED QUEUING NETWORKS WITH SYNCHRONIZATION STATIONS

A PARAMETRIC DECOMPOSITION BASED APPROACH FOR MULTI-CLASS CLOSED QUEUING NETWORKS WITH SYNCHRONIZATION STATIONS A PARAMETRIC DECOMPOSITION BASED APPROACH FOR MULTI-CLASS CLOSED QUEUING NETWORKS WITH SYNCHRONIZATION STATIONS Kumar Satyam and Ananth Krishnamurthy Department of Decision Sciences and Engineering Systems,

More information

Distributed Optimization. Song Chong EE, KAIST

Distributed Optimization. Song Chong EE, KAIST Distributed Optimization Song Chong EE, KAIST songchong@kaist.edu Dynamic Programming for Path Planning A path-planning problem consists of a weighted directed graph with a set of n nodes N, directed links

More information

Online Supplement: Managing Hospital Inpatient Bed Capacity through Partitioning Care into Focused Wings

Online Supplement: Managing Hospital Inpatient Bed Capacity through Partitioning Care into Focused Wings Online Supplement: Managing Hospital Inpatient Bed Capacity through Partitioning Care into Focused Wings Thomas J. Best, Burhaneddin Sandıkçı, Donald D. Eisenstein University of Chicago Booth School of

More information

Semidefinite and Second Order Cone Programming Seminar Fall 2012 Project: Robust Optimization and its Application of Robust Portfolio Optimization

Semidefinite and Second Order Cone Programming Seminar Fall 2012 Project: Robust Optimization and its Application of Robust Portfolio Optimization Semidefinite and Second Order Cone Programming Seminar Fall 2012 Project: Robust Optimization and its Application of Robust Portfolio Optimization Instructor: Farid Alizadeh Author: Ai Kagawa 12/12/2012

More information

On the Convolution Order with Reliability Applications

On the Convolution Order with Reliability Applications Applied Mathematical Sciences, Vol. 3, 2009, no. 16, 767-778 On the Convolution Order with Reliability Applications A. Alzaid and M. Kayid King Saud University, College of Science Dept. of Statistics and

More information

CONTEMPORARY multistation manufacturing systems,

CONTEMPORARY multistation manufacturing systems, IEEE TRANSACTIONS ON AUTOMATION SCIENCE AND ENGINEERING, VOL. 3, NO. 4, OCTOBER 2006 407 Stream-of-Variation (SoV)-Based Measurement Scheme Analysis in Multistation Machining Systems Dragan Djurdjanovic

More information

THE information capacity is one of the most important

THE information capacity is one of the most important 256 IEEE TRANSACTIONS ON INFORMATION THEORY, VOL. 44, NO. 1, JANUARY 1998 Capacity of Two-Layer Feedforward Neural Networks with Binary Weights Chuanyi Ji, Member, IEEE, Demetri Psaltis, Senior Member,

More information

21 Markov Decision Processes

21 Markov Decision Processes 2 Markov Decision Processes Chapter 6 introduced Markov chains and their analysis. Most of the chapter was devoted to discrete time Markov chains, i.e., Markov chains that are observed only at discrete

More information

Agreement algorithms for synchronization of clocks in nodes of stochastic networks

Agreement algorithms for synchronization of clocks in nodes of stochastic networks UDC 519.248: 62 192 Agreement algorithms for synchronization of clocks in nodes of stochastic networks L. Manita, A. Manita National Research University Higher School of Economics, Moscow Institute of

More information

OPTIMAL CONTROL OF A FLEXIBLE SERVER

OPTIMAL CONTROL OF A FLEXIBLE SERVER Adv. Appl. Prob. 36, 139 170 (2004) Printed in Northern Ireland Applied Probability Trust 2004 OPTIMAL CONTROL OF A FLEXIBLE SERVER HYUN-SOO AHN, University of California, Berkeley IZAK DUENYAS, University

More information

Chapter One. Introduction

Chapter One. Introduction Chapter One Introduction With the ever-increasing influence of mathematical modeling and engineering on biological, social, and medical sciences, it is not surprising that dynamical system theory has played

More information

Figure 10.1: Recording when the event E occurs

Figure 10.1: Recording when the event E occurs 10 Poisson Processes Let T R be an interval. A family of random variables {X(t) ; t T} is called a continuous time stochastic process. We often consider T = [0, 1] and T = [0, ). As X(t) is a random variable

More information

WAITING-TIME DISTRIBUTION FOR THE r th OCCURRENCE OF A COMPOUND PATTERN IN HIGHER-ORDER MARKOVIAN SEQUENCES

WAITING-TIME DISTRIBUTION FOR THE r th OCCURRENCE OF A COMPOUND PATTERN IN HIGHER-ORDER MARKOVIAN SEQUENCES WAITING-TIME DISTRIBUTION FOR THE r th OCCURRENCE OF A COMPOUND PATTERN IN HIGHER-ORDER MARKOVIAN SEQUENCES Donald E. K. Martin 1 and John A. D. Aston 2 1 Mathematics Department, Howard University, Washington,

More information

Overload Analysis of the PH/PH/1/K Queue and the Queue of M/G/1/K Type with Very Large K

Overload Analysis of the PH/PH/1/K Queue and the Queue of M/G/1/K Type with Very Large K Overload Analysis of the PH/PH/1/K Queue and the Queue of M/G/1/K Type with Very Large K Attahiru Sule Alfa Department of Mechanical and Industrial Engineering University of Manitoba Winnipeg, Manitoba

More information

September Math Course: First Order Derivative

September Math Course: First Order Derivative September Math Course: First Order Derivative Arina Nikandrova Functions Function y = f (x), where x is either be a scalar or a vector of several variables (x,..., x n ), can be thought of as a rule which

More information

MIT Manufacturing Systems Analysis Lectures 15 16: Assembly/Disassembly Systems

MIT Manufacturing Systems Analysis Lectures 15 16: Assembly/Disassembly Systems 2.852 Manufacturing Systems Analysis 1/41 Copyright 2010 c Stanley B. Gershwin. MIT 2.852 Manufacturing Systems Analysis Lectures 15 16: Assembly/Disassembly Systems Stanley B. Gershwin http://web.mit.edu/manuf-sys

More information