Research On Artificial Immune System And Its Applications In Chemical Engineering

In the 21st century, with the increase of products competition, and the increase of stringent specification of environmental protection, the optimization is an effective technique to reduce the cost and promote the productivity for chemical companies under these pressures. From the product design to the supply chain management, the optimization technique can be used in any scale of chemical process. It is well-known that in chemical engineering, the traditional optimization methods can not be perfect to solve some issues, which are critical and often-encountered, particularly the issues of high nonlinear programming and combinatorial optimization. With the emergence and development of stochastic optimization methods, it is found that these methods are powerful tool to solve these issues. As a branch of stochastic optimization methods, artificial immune system has attracted the attention of the researchers. But there are few reports about its application to chemical engineering, because it emerges too late. In this paper we focus on the research of artificial immune system and its application to chemical engineering.Firstly, we present some optimization methods, such as the traditional optimization methods, some stochastic optimization methods, and one deterministic global optimization method. Secondly, developments, statuses and future research trends of artificial immune system are introduced. Thirdly, according to the future of the issue of phase equilibrium calculation, the issue of dynamic optimization and the issue of scheduling of batch plant, novel approaches are created with improving some artificial immune systems. These approaches are tested by the practical problem. The optimization result and the analysis of principle show that these approaches have better calculation performance. These novel methods not only maintain the diversity of population better, but also find the local optimal point fast. It means that they have better global exploring and local exploiting. Finally, we outline our reach works and future research trends.Our works and contributions are summarized as follows.[1] The calculation of phase equilibrium is not only the base issue in the simulation of chemical process, but also the basic of design and optimization of equilibrium separation system. The calculation of the flash, distillation column, liquid-liquid extraction equipment, and absorber involves phase equilibrium without chemical reaction. The simulation of reactive distillation column, reactive extraction equipment and the reactive absorber involves phase equilibrium with chemical reaction. The classical methods have more limitations. The more complex the system is, the more poor the performance of these methods are. It's well-known that when one system reaches the equilibrium, the value of total Gibbs free energy is minimal. So the issue of calculation of phase equilibrium can be transformed into one problem of optimization. The future of this optimization problem is that only the global solution response to the equilibrium state in fact. So the optimization method to solve this problem must have strong capacity of global searching. A novel approach, Hybrid Immune Algorithm of Antibody Inhibition (Hybrid IAAI, HIAAI), is proposed by means of employing the feasible solution method to deal with the constraints of linear equations, using the convex combination to mutate, and introducing a local exploiting operator (steps quadratic programming) to improve it's capacity of local exploiting. Based on the basic feasible solution, the feasible antibodies are created initially. HIAAI is tested by some benchmarks, and the experimental result show that HIAAI have better performance of global searching, and Its on-line and off-line performance have great improvement. This novel approach is tested by four problems of phase equilibrium without chemical reaction and two problems of phase equilibrium with chemical reaction. The experimental result shows that this approach can find the global solution with high probability and need less effort. Its total performance is better than the Sequential Quadratic Programming (SQP), which is widespread in engineering, and Immune Algorithm of Antibody Inhibition (IAAI).[2] The chemical plants or the bio-chemical plants consist of reaction systems, separation systems, heat exchanger networks and public utilities. The reaction systems and separation systems are the core, accounting for about eighty percent to ninety percent of the investment capitals and operating costs in the total plant. The dynamic optimization of reactors, such as Continuous Stirred Tank Reactor, Plug Flow Reactors and batch reactors, is described as linear or nonlinear differential equations. Then the dynamic optimization method is used to solve these equations to get the optimal control strategy in order to implement the full capacity of the reactor and further to implement the maximal benefits of the plants. The future of these dynamic optimization issues, compared to the common nonlinear programming, is that the time cost to evaluate the objective function once is expensive. There is one kind of issue. In one space its objective function and constraints equation show almost plain, in the other space its objective function and constraints equation show summit drastically. So we employ the Direct Method to solve it. A novel approach, Hybrid OPTimal Artificial Immune NETwork (Hybrid Opt-AiNet, HOpt-AiNet), is proposed by means of employing the operator of local search, such as steps Powell direct searching. This approach has both the ability of the global exploring from OPTimal Artificial Immune NETwork (Opt-AiNet) and the ability of the local exploiting from local optimization algorithms. When infeasible cell appears in the evolution process, the repairing strategy will converse this infeasible cell to feasible cell. The operator mutation is convex combination. When the resolution of time is very high, the sequential refining strategy is used, which reduces the cost time largely. When the parameter of the issue has minor change, the second response can be used, which reduce the time cost abruptly. The mechanism of the second response is that storing the best control strategy ,and considering this control strategy as initial antigen, then creating some initial antibodies by Calling the AiNet. This novel approach is tested by five dynamic optimization examples. The result shows that the approach, which employed the sequential refining strategies, can find the global solution with high probability and just need less effort. Its total performance is better than the Opt-aiNet and some novel approaches from literature. The second response is effective.[3] The future of the dynamic optimization problem with state variable constraints is that making a decision whether the control strategy is feasible or infeasible, will consume time tremendously for getting the value of state variables and constraints need solving the differential equations by the numerical method. For some issues, the minimal disturbance of the control variables causes large changes in the state variables. The method of static penal function is often used to deal with the dynamic optimization problem with state variable constraints. Its shortage is that the size of penal factor has great influence to the running result, and to set the good value of the penal factor is difficult. A novel approach, improved clonal selection algorithm (ICSA), is proposed by means of introducing the immune-network self-study operator, the method of uniform design, and the separation of objective and constraints to deal with the state variable constraints. These methods can improve the capacity of local exploiting, enhance the distribution of the initial population more even, and make the method of handling the constraint more reasonable respectively. This approach is tested by two dynamic optimization problem of reactor. The result shows that this approach can find the global solution with high probability and need less effort. Its total performance is better than some CSA and a few novel approaches in the literature. ICSA does deal with the constraints better, maintain the diversity of population better, and improve the velocity of searching the optimal solution, and the on-line and off-line performance.[4] Batch processes are often used in chemical, pharmaceutical, food and paint industries. The Multi-product Multi-stage and Multi-machine batch plants Schedule under Zero-wait policy, MMMSZ, is one kind of important production mode in the plant. When the production plan is established, the goal of production schedule is to optimize some economic objectives, such as the completion time, and the cost of operation, by getting the processing order or the starting time, and the processing volume of products in the equipment. It's NP-hard. The traditional methods are to transform it into mathematic programming, such as the Mixed-Integer NonLinear Programming (MINLP), and Mixed-Integer Linear Programming (MILP). But these methods are just applied to small-scale practical issues for the number of decision variables and constraints will exponentially increase with the increasing size of the issues. A novel approach, Multi-product Clonal Selection Algorithm (MCSA) is proposed by introducing a novel and efficient strategy to deal with the constraints. Firstly, the better production programs, production batch and entire product batches are created from the production plan. Secondly, these entire product batches are considered as the antibodies, where the gene of antibody is production program. Finally, the optimal antibody is created in the evolution process of clonal selection. For the large-scale schedule issues, Large-scale Multi-product Clonal Selection Algorithm (LMCSA) is created by employing the periodic scheduling strategy. These two approaches are tested by two examples. The result shows that for medium-scale and small-scale problems, MCSA can find the optimal solution within a reasonable time. For large-scale problem, LMCSA can find the optimal solution within a reasonable time, and conquer the dimension disaster.