ELSEVER Computers and Chemical Engineering 24 (2000) 1529-1533 Computers & Chemical Engineering www.elsevier.com/locate/compchemeng Computer-integrated tools for batch process development Jinsong Zhao, Shankar Viswanathan, Chunhua Zhao, Fangping Mu, Venkat Venkatasubramanian * Laboratory for ntelligent Process Systems, School of Chemical Engineering, Purdue University, West Lafayette, CA N 47907, USA Abstract n the current environment of intense market competition, batch process industries stand to benefit from faster process development. Operating procedure synthesis (OPS) and process hazards analysis (PHA) are two time-consuming areas in batch process development because they are often manually performed. Recently, two intelligent systems -- itops and Batch HAZOPExpert (BHE) were developed in our research group to automate OPS and PHA. n this paper, the architecture of the full integration of the two systems is presented. Two applications from specialty chemical industry are presented to demonstrate the utility of the integrated system. 2000 Elsevier Science Ltd. All rights reserved. Keywords: Process hazards analysis; PHA; HAZOP; Expert system; Batch processes 1. ntroduction The increasing trend towards the production of higher-value-added products by specialty chemical, pharmaceutical or agrochemical industries has stimulated considerable interest in batch processes. n the current environment of intense market competition, batch process industries stand to benefit from faster process development. Fast process development provides a competitive advantage by facilitating fast market penetration and exploiting patent protection. The current trend (Allgor, Barrera, Barton & Evans, 1996) in the specialty chemicals industry is toward the manufacture of products with shorter life cycles and higher functionality that are tailored to specific market niches. Thus, new products are introduced very frequently. Due to increased public concerns and strict regulations of federal laws on about occupational safety, there have already been major changes concerning safety management in chemical process industries (CP). These changes include the widespread use of systematic process hazards analysis approaches such as HAZOP (Hazard and Operability) analysis, Check-list, What-if analysis, and the increased number of safety engineers * Corresponding author. Tel. + 1-765-4940734; fax: + 1-765- 4940805. E-mail address: venkat@ecn.purdue.edu (V. Venkatasubramanian). and their abilities. n batch process development, however, operating procedure synthesis (OPS) and process hazards analysis (PHA) take considerable amount of time and effort because they are often manually performed. Thus, there exists substantial motivation to automate OPS and PHA for batch processes by developing computer-based approaches. Recently, two intelligent systems for OPS and HAZOP of batch processes have been developed in our research group. Viswanathan, Johnsson, Srinivasan, Venkatasubramanian and Arzen (1998a,b) presented an intelligent tool for operating procedure synthesis (itops) by using grafchart-based methods starting from the chemist's process description. Srinivasan and Venkatasubramanian (1998) developed an automated HAZOP analysis expert system -- Batch HAZOPExpert (BHE) for batch processes. Viswanathan, Zhao and Venkatasubramanian (1999) presented a preliminary framework for integrating itops and BHE. According to the framework, BHE could automatically find out all of the potential hazards caused by all possible deviations to the process variables based on the information in the operating procedure generated by itops. Since the operating procedure is used to instruct an operator to safely and optimally manage the batch process, safety issues related to each operation should be explicitly flagged in the corresponding operation instructions. However, the preliminary framework 0098-1354/00/$ - see front matter 2000 Elsevier Science Ltd. All rights reserved. P: S0098-1354(00)00559-7 转载
1530 J. Zhao et al./ Computers and Chemical Engineering 24 (2000) 1529-1533 did not provide an internal path to feed the safety hazards reported by BHE back into the operating procedures generated by itops. The feedback of the safety information had to be manually done. To automatically feed back the safety information into the operation instructions, a fully integrated system of itops and BHE will be addressed in this paper. ndustrial applications of this integrated system will be focused in specialty chemical processes. 2. ntegrated system of itops and BHE Fig. 1 shows the architecture of the integration system of itops and BHE. Material and equipment databases are combined within the integration system to obtain quantitative hazard critical properties of process materials and design parameters of process equipment. Material interaction database is used to capture hazards caused by possible side reactions. Block Process Sequence Diagram (Block PSD) is a high-level process description. t represents the sequence of the main tasks of one batch process, itops provides a user-friendly interface to specify manually the Block PSD based on the chemist's process description. The input to the integrated system is just the process-specific information including process materials, process equipment, process chemistry (reactions and separations) and Block PSD. Once the process-specific information is specified, process sequence diagram (PSD) and detailed batch operation instructions are automatically generated by itops. PSD graphically indicates the sequence of the operations. BHE represents batch processes by two-level Petri-Nets. The top level is called Recipe Petri Net (RPN) indicating the sequence of the main tasks. Associated with each task, there is a Task Petri Net (TPN) indicating the sequence of the necessary subtasks. Associated with each subtask, there is a digraph model qualitatively capturing the relationships between the process variables of the subtask. Details of itops and BHE can be found in literature (Srinivasan & Venkatasubramanian, 1998; Viswanathan et al., 1998a,b). The OPS-PHA interface converts Block PSD into RPN, and PSD into TPNs. As the databases can not provide all the required hazard critical properties of the process materials and equipment, additional information such as the toxic nature of a material has to be put in by users through user-friendly interfaces according to the instructions given by the integrated system. Once the additional information required by the system is specified, HAZOP analysis of the whole process can be automatically performed. The digraph model library covers models of about forty operations that are often used in batch processes. The knowledge base containing rules for utility failure analysis was added to BHE recently to enhance the capability in regular HAZOP analysis. As TPNs are converted from the PSD, mapping pointers are used to associate subtasks with the corresponding operations. Safety hazards found in a particular subtask can automatically be added to the batch operation instructions of the operation associated with the subtask. That helps the operator be aware of the corresponding safety hazards and avoid maloperations. 3. ndustrial applications The following two examples come from two major specialty chemicals manufacturers. The identities of the compounds involved have been concealed. Example 1 Fig. 2 is the state task network for the industrial example reported by Allgor et al. (1996). According to the process description of the chemists, the process is to t --Reactions & Separations ~ w, L. ' L" i,] Fig. 1. Architecture of the integrated system.
J. Zhao et al. / Computers and Chemical Engineering 24 (2000) 1529-1533 1531 H? ) ChKi~RS(~L [ for 0.5 hour t, f~ 0.25 hour / t Fig. 2. State task network for example 1 (Allgor et al., 1996). manufacture a fixed quantity of a monomer. Since the development of similar products by competitors is imminent, the process development is subject to a strict time horizon constraint. Hence rapid development of an efficient process is critical to the success of launching the new product. According to the chemist's process description, to manufacture the monomer, six tasks including two exothermic reactions, one endothermic reaction, two atmospheric distillations and one vacuum distillation were required. Totally, there were 14 process materials including reactants, by-products, solvents and catalysts, and seven pieces of process equipment including reactors, receivers and distillation columns. The PSD generated by the integrated system contains 34 operations including charge, purge, heat, cool, transfer, hold, distillation and so on. Fig. 3 shows the PSD of Task-1. Two hundred and seventy potential process variable deviations were analyzed, and 38 safety hazards were captured for these deviations. t took the integrated [ for 0.5 hour t!.lun'y (J movl) 1 ~ o.~ l~r / [ c~r2z~0t ] fo 0.5 hour Fig. 3. PSD of Task 1 for Example 1. system about 5 s to complete the HAZOP analysis. Table 1 lists the batch operation instructions of the first Table 1 Modified batch operation instructions of Task-1 in Example-1 Number Batch operation instructions Potential hazards lo 11 12 13 Charge 5001 R2 to reactor-1 Charge 5001 toluene to reactor-1 for 0.5 h Heat reactor-1 to 70 C for 0.5 h Load catalyst slurry (5 mol/1) 1 1 Charge R2 2501 to reactor- Load catalyst slurry (5 mol/l) 1 1 to reactor-1 Hold for 2 h Heat reactor-1 to 90 C for 2 h Take a sample of the product from reactorl for purity acceptance testing Cool to 25 C for 0.25 h High temperature can lead to vaporization of volatile materials. Low agitation leads to poor mixing. High temperature leads to potential fire hazard. High charge flow rate leads to static charge. Low agitation leads to poor mixing. High temperature leads to potential fire hazard. Short time leads to incomplete reaction. High temperature can lead to decomposition of A and R2, high gas generation rate of H and swelling reactor content. Operator exposure to hot and hazardous materials during sampling.
1532 J. Zhao et al./ Computers and Chemical Engineering 24 (2000) 1529-1533 ~wca4 ToPK~ To ~m2 i Fig. 4. Flow chart of Example 2. Transfer J Strip storage [ Tra"sfer stooge [ ] Tr~sf~, [ l Transfer [ Fig. 5. Block PSD of Example 2. task containing the safety analysis results. Due to the length limitation, the batch operation instructions of other tasks are not presented here. Example 2 This example addresses the OPS and PHA of a batch process with material transportation in pipelines and storage in tanks. The purpose of storage is to smooth fluctuation in the flows in and out. The risks from the process industries arise from processes, storage and transport. The historical record shows that storage and transport are two major contributors (Lees, 1996). The disaster in Union Carbide's Sevin plant, Bhopal, ndia in 1984 just resulted from toxic vapor cloud in a storage tank. The PHA in storage and transport is, therefore, extremely important. Fig. 4 shows the flow sheet of this process. The process material is a kind of monomer which is highly T,,,a, ~ T~t.l/2 ~'~" ~ "~;' '--" r,-~,-bl 'l'ult.~l -~.~- ~ ~-,-~,.-,t----------~ r-.,,,~,~,i Fig. 6. PSD of Example 2.
J. Zhao et al./ Computers and Chemical Engineering 24 (2000) 1529-1533 1533 Table 2 Comparison of HAZOP results of Task-3 in Example-2 Deviations Hazards from the Hazards from the team integrated system High Tank overpressure, Potential tank temperature rupture overpressure leading to rupture Potential fire hazard High pressure Overpressure Potential tank floating roof and overpressure leading to tank rupture Vacuum Tank rupture Potential tank rupture pressure under vacuum pressure High level Loss containment of Potential overflow leading monomer to loss of containment Low level None ntent of operation not met hazardous and flammable. The monomer is distributed by the pump P-1 from the source tank to two big storage tanks (Storage Tank-1 and Tank-2). Then, the Strip Pump is started to strip the monomer in the pipes back to a receiver tank. From Storage Tank-1 and Tank-2 to Storage Tank-3 and Tank-4, the monomer is transported by pumps (P3 and P2) through a long distance of pipelines. Then the monomer in Storage Tank-3 is sent to Plant-1 while the monomer in Storage Tank-4 is sent to Plant-2. The Block PSD input to the integrated system is shown in Fig. 5. The PSD generated by the system is shown in Fig. 6. The PHA team reported 34 safety hazards by analyzing 41 process variable deviations. Based on the process information in PSD, 45 process variable deviations were analyzed by the integrated system. t took the system about 3 s to finish the HAZOP analysis. The HAZOP report, then, is immediately available in a Microsoft Word file in the required format. Out of the 34 safety hazards reported by the PHA team, only one safety hazard was not captured by the integrated system because that hazard was very process specific. n addition to these hazards, 71 more hazards were flagged by the integrated system including some potential fire hazards. Table 2 shows the comparison of the Task-3 (storage)'s safety hazards obtained from the PHA team and the integration system. t can be seen that potential fire hazard and low storage level was neglected by the team. Flagging these hazards will not only be helpful for the process engineers to think of necessary safeguards, but also be helpful for the operators to monitor the temperature and level in the storage tank during the related operations. 4. Conclusions The results of the application of the integrated system to two specialty chemicals industry processes are presented. The intelligent system considers many more potential hazardous scenarios than the PHA team does. Therefore, a more comprehensive and consistent PHA can be performed by the system. The batch operation instructions that have the captured safety hazards associated with them will help operators manage the batch processes more safely. Meanwhile, documentation of PSD, batch operation instructions and HAZOP analysis report becomes much easier, especially when changes are required in chemist's process description. References Allgor, R. J., Barrera, M. D., Barton, P.., & Evans, L. B (1996). Optimal batch process development. Computers & Chemical Engineering, 20, 885 896. Lees, F. P. (1996). Loss prevention in the process industries: hazard identification, assessment, and control, vol. 2 (2nd ed.). London: Butterworth-Heinemann. Srinivasan, R., & Venkatasubramanian, V. (1998). Automating HAZOP analysis of batch chemical plants: part algorithms and application. Computers & Chemical Engineering, 22, 1357-1370. Viswanathan, S., Johnsson, C., Srinivasan, R., Venkatasubramanian, V., & Arzen, K. (1998a). Automating operating procedure synthesis for batch processes -- part knowledge representation and planning framework. Computers & Chemical Engineering, 22, 1673 1685. Viswanathan, S., Johnsson, C., Srinivasan, R., Venkatasubramanian, V., & Arzen, K. (1998b). Automating operating procedure synthesis for batch processes -- part implementation and application. Computers & Chemical Engineering, 22, 1687-1698. Viswanathan, S., Zhao, J., & Venkatasubramanian, V. (1999). ntegrating operating procedure synthesis and hazards analysis automation tools for batch processes. Computers & Chemical Engineering, 23, $747-$750.