Manojkumar Ramteke

Optimal variable selection for effective statistical process monitoring

Abstract In a typical large-scale chemical process, hundreds of variables are measured. Since statistical process monitoring techniques typically involve dimensionality reduction, all measured variables are often provided as input without weeding out variables. Here, we demonstrate that incorporating measured variables that do not provide any additional information about faults degrades monitoring performance. We propose a stochastic optimization-based method to identify an optimal subset of measured variables for process monitoring. The benefits of the reduced monitoring model in terms of improved false alarm rate, missed detection rate, and detection delay is demonstrated through PCA based monitoring of the benchmark Tennessee Eastman Challenge problem.

Unprecedented adsorptive removal of Cr2O72− and methyl orange by using a low surface area organosilica

An extraordinary adsorption capacity of 359 and 1679 mg g−1 for the adsorptive removal of Cr2O72− and methyl orange (MO), respectively, was observed by using a low surface area (SABET of 10 m2 g−1) organosilica. The organosilica samples were synthesized by condensing phosphonitrilic chloride trimer with N1-3-(trimethoxysilylpropyl)ethylenetriamine followed by hydrolysis and co-condensation with different amounts of tetraethyl orthosilicate (TEOS). The kinetics, isotherms and thermodynamics of the adsorption process were investigated. The very high adsorption capacity of the low surface area adsorbent could be attributed to the formation of hemimicelles at the adsorbed sites of the adsorbent which suddenly increases the adsorption rate. An isotherm model was proposed similar to the adsorption model reported by Zhu and Gu with some modification. A pseudo-second order kinetic model fits best with the experimental data. Thermodynamically, as expected the present adsorption process is spontaneous and exothermic.

Simulated binary jumping gene

A jumping gene operator, simulated binary jumping genes (SBJG) is developed for real-coded NSGA-II.Performance of SBJG is measured by calculating performance metrics for 37 test problems.SBJG simulates the effect of binary-coded jumping gene operator effectively.SBJG achieves faster convergence in lower number of generations compared to other JG operations and real-coded NSGA-II.SBJG performed well for multi-objective optimization of industrial steam reformer. The concept of jumping gene from biology has become quite popular for increasing the convergence speed of binary-coded elitist non-dominated sorting genetic algorithm. This inspired several researchers to implement this concept in real-coded elitist non-dominated sorting genetic algorithm which is free from limitations of binary coding. However, these implementations have achieved only a limited success. This is primarily due to their focus on mimicking the procedure instead of simulating its effect whereas the latter suits more to the real nature of variables as simulated forms of the crossover and the mutation operations are commonly used in real-coded genetic algorithm. In order to address this shortcoming, a new jumping gene operator, namely, simulated binary jumping gene is developed in the present study. For this, a detailed qualitative analysis of all existing jumping gene operators is performed. Unlike other real-coded jumping gene operators, the new operator simulates the concept of jumping gene closely to that used in the binary version. The efficacy of the new operator is then tested quantitatively using well-known indicators of generational distance, hyper-volume ratio and spacing over thirty-seven challenging multiobjective optimization problems from the literature. The results obtained with the inclusion of newly developed operator show a significant increase in convergence speed of real-coded elitist non-dominated sorting genetic algorithm, particularly for the restricted number of generations. Also, the performance of the algorithm with the new operator is found to be better than that with other existing real-coded jumping gene operators. The effectiveness of the new operator in achieving faster convergence for real-life multi-objective optimization problems is further established by solving the industrial problem of multiobjective optimization of a dynamic steam reformer.

Novel genetic algorithm for short-term scheduling of sequence dependent changeovers in multiproduct polymer plants

Abstract Polymer plants generally operate to produce different grades of product from the same reactor. Such systems commonly require short-term scheduling to meet market demand. One important requirement in continuous-time scheduling of such systems is to satisfy a variety of constraints, including identifying feasible sequences of the predecessor and successor jobs to effectively handle changeovers. In this study, a new genetic algorithm (GA) is proposed to solve such job sequencing problems. The proposed GA uses real-coded chromosome to represent job orders and their sequences in the schedule. The novelty is that the representation ensures that all constraints are satisfied a priori, except the sequence constraint which is handled by penalizing violations. Three important problems relevant to polymer industry are solved to obtain optimal schedules. The first deals with the sequencing constraint between individual product orders, the second with sequencing constraint between groups of product orders, while the third incorporates batching with scheduling.

An elitist non-dominated sorting bat algorithm NSBAT-II for multi-objective optimization of phthalic anhydride reactor

Abstract Nature inspired meta-heuristic algorithms are an integral part of modern optimization techniques. One such algorithm is bat algorithm which is inspired from echolocation behavior of bats and has been successfully applied to non-linear single-objective optimization problems. In this paper, a multi-objective extension of bat algorithm is proposed using the concepts of Pareto non-dominance and elitism. The novel algorithm is tested using thirty multi-objective test problems. The performance is measured using metrics namely, hyper-volume ratio, generational distance and spacing. The newly developed algorithm is then applied to a real-world multi-objective optimization problem of a phthalic anhydride reactor. It shows faster convergence for test problems as well as the industrial optimization problem than two popular nature inspired meta-heuristic algorithms, i.e. multi-objective non-dominated sorting particle swarm optimization and real-coded elitist non-dominated sorting genetic algorithm.

Applications of Genetic Algorithms in Chemical Engineering II: Case Studies

Chemical engineering systems encompass a wide variety of processes from the production of bulk chemicals to highly sophisticated specialty chemicals, their purifications, control, planning, and scheduling. Optimization is an important operation to refine the complex decision-making of these systems. Recently, genetic algorithm (GA) has been used extensively for such applications due to its ability to handle multiple objectives simultaneously. Many new adaptations of GA have been developed over the last two decades to facilitate the solution of such complex problems. The rapidly improving computational speeds further bolsters the effective utilization of these algorithms. A few interesting studies are the optimization of polymerization reactors, catalytic reactors, separations equipment, planning and scheduling, combinatorial problems, and data-driven applications. These studies give interesting insights to improve the operations and also to pave the way to handle other complex interdisciplinary systems.

Applications of Genetic Algorithms in Chemical Engineering I: Methodology

The fascinating world of genes has been an inspiration for mankind. One such inspiration has led to a popular optimization technique, genetic algorithm (GA). Its inherent parallelism has enabled significant computational improvement over deterministic enumerations. Further, it has provided a flexibility of solving multiple objectives in a derivative-free environment. These advantages are extremely useful for solving optimization problems in chemical engineering, ranging over a wide variety of processes from the production of bulk chemicals to highly sophisticated specialty chemicals, their purification, control, planning, and scheduling. These systems are often associated with multiple objectives and complex model equations. Several variations of GA have been developed over the last four decades by incorporating ground-breaking concepts such as elitism, jumping gene, crowding distance, ranking, altruism, etc., to enable faster convergence of the algorithms. Continuous improvements are being made by the use of new or hybrid concepts so as to provide improved applicability and flexibility, and so as to exploit the rapidly increasing computational speeds.

Integrating Graph-based Representation and Genetic Algorithm for Large-Scale Optimization: Refinery Crude Oil Scheduling

Abstract Scheduling optimization problems are often associated with large number of variables and combinatorial constraints. These problems can be represented graphically through a network structure. This graphical representation can provide important insights to handle the combinatorial constraints. In this study, the graphical representation is incorporated in the framework of genetic algorithm to solve large-scale refinery crude oil scheduling problems. Our results show that use of such graphical representation offers significant advantages while solving multi-objective, multi-solution and nonlinear formulations in reasonable computational time.

Reactive Scheduling of Crude Oil using Structure Adapted Genetic Algorithm under Multiple Uncertainties

Abstract Crude oil processed in marine access refineries contributes about 15% of the total energy production worldwide. An optimized schedule of crude unloading and charging in these offers the best utilization of available resources to increase the profitability and also helps in incorporating the future uncertainties commonly encountered in the operation. In the present study, a new reactive crude oil scheduling methodology is developed for marine-access refinery using a structured adapted genetic algorithm to handle the commonly encountered uncertainties of increase in demand and ship arrival delay. Three different industrial examples with 21, 21 and 42 periods are solved for above uncertainties with single and multiple objectives. In the single-objective formulation, profit is maximized whereas in multi-objective formulation an additional objective of inter-period deviation in crude flow to distillation units is minimized. The results obtained show the efficient handling of uncertainties with improved profitability and operability of the plant.

Optimized on-line control of MMA polymerization using fast multi-objective DE

ABSTRACT Optimized on-line control (OOC) of polymerization reactors combine the optimization with the on-line operation and control. In this, re-optimized control variable trajectories, in the presence of unplanned disturbances, are obtained and implemented on-line to save the batch. Also, the available computational time for the optimization is limited as the re-optimized trajectories need to be implemented in real time on the actual system. In the present study, the OOC of such a system, i.e., bulk polymerization of methyl methacrylate (MMA) in a batch reactor, is carried out in the occurrence of heater malfunction. To solve the underlying multi-objective problem, a multi-objective variant of differential evolution with an improved mutation strategy is developed. The developed algorithm shows faster convergence with respect to other compared algorithms for a large number of benchmark problems. Finally, this algorithm is used to find the optimal temperature trajectories and the OOC with these trajectories found to be successfully countering the effect of heater malfunction.