International Journal of Application or Innovation in Engineering & Management (IJAI (IJAIE AIEM) Web Site: www.ijaiem.org Email: editor@ijaiem.org ISSN 2319 - 4847 Special Issue for International Technological Conference-2014 Optimization of PID Parameters using Dual Population Genetic Algorithm Dr. Shanta Sondur1, Anil Kumar R. Sudele2 1 Information Technology Dept., VES Institute of Technology, Mumbai, India shantasondur@gmail.com 2 CRG Engineer, SMEC Automation Pvt. Ltd., Mumbai, India anilsudele@gmail.com ABSTRACT The existing Dual Population Adaptive Genetic Algorithm (DPAGA) requires more number of iterations and longer time duration for optimization. A modification has been suggested for overcoming the same. In the existing algorithm the adaptive crossover and adaptive mutation is done in the main population and immediately the convergence condition is checked by evaluating the individuals of the main population only whereas, in the proposed algorithm, the adaptive crossover and adaptive mutation are performed for both the main population and the subordinate population. Then the individuals of both the populations are evaluated for checking the convergence condition. This improves the efficiency of the proposed algorithm and it gives solution in lesser iterations. The existing DPAGA and proposed algorithm has been applied for optimization of the PID parameters for some Plant Transfer Functions. The performance of both the methods of optimization of PID parameters has been compared. Keywords: DPAGA, PID parameters, Optimization, Adaptive Mutation, Adaptive Crossover etc. 1. INTRODUCTION The Dual-Population Genetic Algorithm (DPGA) was originally proposed for stationary optimization problems [3, 4]. It consists of two distinct populations with different evolutionary objectives. In DPGA [5], the main population plays the same role as that of the population of an ordinary Genetic Algorithm. It evolves to find a good solution with a high fitness value. The additional population called the reserve population provides additional diversity to the main population. In order to allow the main population to use the diversity in the reserve population, there must be a method of exchanging information between the populations. The migration method used for most multi-population GAs (MPGAs), however, is not suitable for DPGA because the two populations have different evolutionary objectives. An individual of one population can hardly survive in the other because the methods for evaluating fitness are different for both populations. Therefore, DPGA uses crossbreeding as a means of information exchange. In DPGA, offspring are produced by mating not only the individuals of the same population but also the individuals of different populations. Since the crossbred offspring contain the genetic material of both populations, their fitness values are not too low and thus they assimilate relatively easily into the new population. Another Dual Population Genetic Algorithm with Evolutionary Density (DPGA-ED) [3] has been introduced which is the improved version of DPGA [5]. In DPGA-ED the reserve population evolves by itself to maintain the diversity whereas that in the DPGA [5] cannot evolve on its own and depends of the genetic material imported from the main population. DPGA-ED shows performance improvements and avoids the premature convergence. The DPGA [6] generalizes DPGAs and provides a more thorough analysis on a wider variety of problem classes using binary, real-valued, and order-based representations. It has been shown that the main population of DPGA can become more exploitative than the population of standard GAs when it approaches a peak, but can also become much more explorative when it tries to escape from a local peak. This change of mode is made possible by the existence of the reserve population that adaptively maintains appropriate distance to the main population and thus provides controlled diversity to the main population. Latest version of Dual Population Adaptive Genetic Algorithm (DPAGA) [7] involves adaptive adjustment of the crossover probabilities and mutation probabilities for both main and the reserved population. The crossover probability and mutation probability get adjusted dynamically in the course of evolution according to the actual situation of population. The DPGA [3] and DPGA-ED [6] take thousands of iterations required for the high dimensional unimodal and multimodal benchmark test functions. Whereas in DPAGA [7], the maximum number of iterations required for finding the solutions for the benchmark test functions was 300. Looking at the slow convergence rates (More number of iterations) we would like to propose some changes to improve the convergence rate and time by optimizing the number of iterations. The objective is to optimize the number of iterations of the algorithm discussed in [7], apply that for the optimization of Organized By: Vivekanand Education Society's Institute Of Technology International Journal of Application or Innovation in Engineering & Management (IJAI (IJAIE AIEM) Web Site: www.ijaiem.org Email: editor@ijaiem.org ISSN 2319 - 4847 Special Issue for International Technological Conference-2014 PID parameters and compare the performance of the existing method with the modified method. 2. EXISTING ALGORITHM Here we describe the existing DPAGA algorithm given in [7]. Genetic algorithm based on adaptive evolution in dual population (DPAGA) [7] is a novel improved adaptive genetic algorithm. In this method, the authors have discussed about the new concept of adaptive mutation and crossover for both the main and the reserved population. The crossover probability and mutation probability get adjusted dynamically in the course of evolution according to the actual situation of population. However the algorithm takes more number of iterations for convergence. Its flowchart is shown in figure 1. Figure 1. The working flow of DPAGA Following are the steps of the existing algorithm. Steps of Algorithm: Algorithm BEGIN • Step 1: Code genetic individuals and initialize population ; • Step 2: Calculate individual fitness; • Step 3: Perform selecting operation according to the roulette wheel selection method, and gain the new population (main population) and the eliminated population (subordinate population); • Step 4: For the main population , perform adaptive crossover operation with high probability and adaptive mutation operation with low probability; • Step 5: Judge convergence condition of . If it is congruous, perform Step 9; • Step 6: For the subordinate population , perform adaptive crossover operation with low probability and adaptive mutation operation with high probability; • Step 7: Form the new population from the results of Step 4 and Step 6; • Step 8: Repeat Step 2 ~ Step 6, till the convergence condition is congruous; • Step 9: Stop. Algorithm END Organized By: Vivekanand Education Society's Institute Of Technology International Journal of Application or Innovation in Engineering & Management (IJAI (IJAIE AIEM) Web Site: www.ijaiem.org Email: editor@ijaiem.org ISSN 2319 - 4847 Special Issue for International Technological Conference-2014 Its population includes a new population formed by selection operation (the main population) and a population eliminated by selection operation (the subordinate population). Dual crossover means the main crossover operator and the subordinate crossover operator . Dual mutation operator means the main mutation operator and the subordinate mutation operator . In the evolution of the main population, the individuals perform adaptive crossover operation with high probability and adaptive mutation operation with low probability according to the main crossover operator and the main mutation operator. And in the evolution of the subordinate population, the individuals perform adaptive crossover operation with low probability and adaptive mutation operation with high probability according to the subordinate crossover operator and the subordinate mutation operator. The dual crossover operator and dual mutation operator are adjusted adaptively according to the individual fitness in the evolution of population. Their formulae can be described as follows: + = , = = 1 + exp + , + , = − , (./0 = ( < ( *+ − 1 + exp !" %& , ( ≥ ( *+ - !" %& , - # $ - !" 1 1 + 1/2.−/0 ( < ( *+ − 1 + exp ( ≥ ( *+ # $ !" ( < ( *+ − 1 + exp + !" %& , # $ !" - !" ( ≥ ( *+ %& , ( ≥ ( *+ - # $ - !" ( < ( *+ (1) (2) (3) (4) (5) Where, ( and ( denotes the maximal fitness of the main population and subordinate population respectively, ( *+ and ( *+ denotes the average fitness of the main population and subordinate population respectively, ( denotes the higher fitness of the two crossing individuals, ( denotes the fitness of the mutating individuals, and denotes the lower limit and the upper limit of the crossover probability of the main population respectively, and denotes the lower limit and the upper limit of the crossover probability of the subordinate population respectively, and denotes the lower limit and the upper limit of the mutation probability of the main population respectively, and denotes the lower limit and the upper limit of the mutation probability of the subordinate population respectively. Generally, their values may be chosen as follows, = 0.6, = 0.9, = 0.001, = 0.1, = 0.001, = 0.1, = 0.1, = 0.5. From formula (1) ~ formula (4), we can see that the probability of crossover and the probability of mutation are adjusted nonlinearly according to the function (./0 between the average fitness and the maximal fitness. When the fitness of the majority of individuals is near and the average fitness is close to the maximal fitness, the probability of crossover and the probability of mutation will be elevated. So, the individuals nearby the maximal fitness are preserved as many as possible. 3. THE PROPOSED ALGORITHM In DPAGA [7] the workflow suggests that the convergence condition should be checked by evaluating the individuals of the individuals of the main population only after the adaptive mutation and crossover have been done. Later the adaptive mutation and crossover are performed in the individuals of the subordinate populations and then are evaluated and ranked. Later on the populations are sorted in to main and subordinate population using roulette wheel selection. Checking the convergence condition before performing crossover, mutation and evaluating the individuals in the subordinate population Organized By: Vivekanand Education Society's Institute Of Technology International Journal of Application or Innovation in Engineering & Management (IJAI (IJAIE AIEM) Web Site: www.ijaiem.org Email: editor@ijaiem.org ISSN 2319 - 4847 Special Issue for International Technological Conference-2014 and considering only the individuals of the main population reduces the chances of selection of the best individual that may have been evolved in the subordinate population. Therefore this algorithm takes many iterations for solving the problems. The modification suggested is to perform crossover and mutation of the individuals of the main population and form a new main population. Then perform crossover and mutation of the subordinate population and form a new subordinate population. Later on convergence condition is checked by evaluating individuals of both the populations i.e., the main population and subordinate population. This can reduce the number of iterations. In terms of steps, we propose to carry out step number 5 after step number 6 and 7. Steps 1, 2, 3 and 4 will remain the same in the proposed algorithm. However, instead of checking the convergence condition at step 5 we prefer it to be done after step 6 and 7. Following are the steps of the proposed algorithm. The flowchart of the proposed DPAGA is shown in Figure 2. The selection method remains the same Roulette Wheel Selection. Also formulae for the mutation and crossover probabilities of both the population remains same as DPAGA [7]. Figure 2. The working flow of the proposed DPAGA Following are the steps of the proposed algorithm. Steps of Algorithm: Algorithm BEGIN • Step 1: Code genetic individuals and initialize population ; • Step 2: Calculate individual fitness; Organized By: Vivekanand Education Society's Institute Of Technology International Journal of Application or Innovation in Engineering & Management (IJAI (IJAIE AIEM) Web Site: www.ijaiem.org Email: editor@ijaiem.org ISSN 2319 - 4847 Special Issue for International Technological Conference-2014 • Step 3: Perform selecting operation according to the roulette wheel selection method, and gain the new population (main population) and the eliminated population (subordinate population); • Step 4: For the main population , perform adaptive crossover operation with high probability and adaptive mutation operation with low probability; • Step 5: For the subordinate population , perform adaptive crossover operation with low probability and adaptive mutation operation with high probability; • Step 6: Form the new population from the results of Step 4 and Step 5; • Step 7: Judge convergence condition of . i.e., by considering the individuals of the main population and subordinate population . If it is congruous, perform Step 9; • Step 8: Repeat Step 2 ~ Step 7, till the convergence condition is congruous; • Step 9: Stop. Algorithm END 4. IMPLEMENTATION AND RESULTS The DPAGA [7] and the Proposed Algorithm have been used for optimization of PID Parameters for various Plant transfer functions. The following plant transfer functions (TFs) were used. 3 .40 = 20 324 8 + 24 + 9 (6) 342.5 + 7.44 + 324.5 1 3> .40 = 4.4 + 20.4 + 40 38 .40 = 48 (7) (8) 400 + 304 8 + 2004 4.228 3@ .40 = 8 .4 + 0.50.4 + 1.644 + 8.4560 3? .40 = 3C .40 = 4> (9) (10) 27 .4 + 10.4 + 30> (11) 1.6 4 8 + 2.5844 + 1 406.14 + 178.9 3E .40 = ? > 4 + 7.14 + 121.94 8 + 71.84 + 21.5 3D .40 = (12) (13) Here, F denotes the population size and 3 denotes the maximum number of Generations. Two trials (Trial 1 and Trail 2) are performed for optimizing the PID Parameters using the DPAGA [7] based PID and Proposed DPAGA based PID. F is taken as 100. Trail 1 was performed till the 30 generations. Trial 2 was performed till there was no significant difference between the best individuals for 5 consecutive generations or till 10,000 generations were reached. 3 denotes the number of iterations required for optimization while, GH , GI and GJ denotes Proportional, Integral and Derivative gains of the PID Controller. The selected domain for the parameters, GH , GI and GJ of the Transfer Functions 3 .40 to 3E .40 for the DPAGA [7] and Proposed Algorithm are shown in Table 1. Table1. The initial domain selected for GH , GI and GJ Plant TFs 3 .40 38 .40 3> .40 3? .40 3@ .40 3C .40 3D .40 3E .40 Range of KL 2 - 20 0 - 120 0 - 50 2-8 1-3 0 - 18 0 - 31 0 -16 Range of KM 0-6 0 - 25 0 - 25 0-5 0-1 0 - 17 0 - 19 0 -11 Range of KN 0-6 0 - 10 0 - 25 0-1 0-2 0 - 10 0 - 19 0-1 Organized By: Vivekanand Education Society's Institute Of Technology International Journal of Application or Innovation in Engineering & Management (IJAI (IJAIE AIEM) Web Site: www.ijaiem.org Email: editor@ijaiem.org ISSN 2319 - 4847 Special Issue for International Technological Conference-2014 The ‘Integral of Absolute Magnitude of the Error’ (IAE) has been used as the evaluation criteria for the individuals of both the populations. Other specifications, such as overshoot, rise time and settling time may also be taken into account. In order to overcome the large energy of the controller, the square term of control output O.P0 is added to the fitness function [9]. The Objective function is U Q = R [T |1.P0| + T8 O8 .P0]YP + T> . PZ (14) Q = ^V [T |1.P0| + T8 O8 .P0 + T? |1[.P0|]YP + T> . PZ , 1[.P0 < 0 (15) V Where, w1, w2, w3 are weight coefficients, O.P0is the output of controller, PZ is rise time and 1.P0 is the system error. In order to get satisfactory transient response and to suppress overshoot, Objective function [10] [11] was revised as follows: U 1[.P0 = [.P0 − [.P − 10 (16) Where T? is the weight coefficient, and T? >>T , [ is the output of the plant. Where there is overshoot, a punishment term 1[.P0 is added at once. Weight coefficients are T = 0.999, T8 = 0.001, T> = 0.2 and T? = 100. The fitness function is chosen as (= 1 Q (17) The fitness Function is the inverse of IAE. The elapsed time and optimum values of GH , GI and GJ parameters obtained using the DPAGA [7] based PID and the Proposed DPAGA based PID for the selected transfer functions in Trial 1 is shown in Table 2 and that for Trial 2 is shown in Table 3. Figure 3 to 6 shows response for the transfer functions 3C .40 and 3E .40 in Trial 1 and trail 2. Table 2. The PID parameters obtained in Trial 1 Trial 1 Plant TFs 3 .40 38 .40 3> .40 3? .40 3@ .40 3C .40 3D .40 3E .40 The Existing DPAGA[7] – PID Time KL KM KN taken in seconds 4.8370 2.7032 1.2465 54.144 5.2043 1.0842 0.4337 54.144 27.3441 16.3570 13.7539 40.860 6.2562 0.9195 0.2373 39.707 1.3737 0.4018 1.8577 39.894 16.2000 15.3000 9.0000 40.268 1.9758 1.2110 1.2110 46.086 14.4000 9.9000 0.9000 49.204 \] 30 30 30 30 30 30 30 30 The Proposed DPAGA[7] – PID Time KL KM KN taken in \] seconds 18.200 5.4000 5.4000 36.697 30 108.00 22.500 9.0000 36.796 30 45.000 22.500 22.500 27.038 30 7.4000 0.5000 0.9000 27.038 30 2.8000 0.9000 1.8000 27.491 30 1.3207 1.2474 0.7337 17.845 30 27.900 17.100 17.100 31.721 30 1.1732 0.8065 0.0733 33.604 30 Table 3. The PID parameters obtained in Trial 2 Trial 2 Plant TFs 3 .40 38 .40 3> .40 3? .40 3@ .40 3C .40 3D .40 3E .40 KL 3.8 12.000 45.000 6.2562 1.2 16.200 3.1000 14.400 The Existing DPAGA[7] – PID Time KM KN taken in seconds 0.6 0.6 1530.2 2.5000 1.0000 18.271 22.500 22.500 19.845 0.9195 0.2373 40.588 0.1 0.2 1398.4 15.300 9.0000 15.405 1.9000 1.9000 106.37 9.9000 0.9000 18.192 \] 1104 14 15 16 1105 9 57 9 The Proposed DPAGA[7] – PID Time KL KM KN taken in \] seconds 18.200 5.4000 5.4000 9.1364 9 108.00 22.500 9.0000 20.557 22 45.000 22.500 22.500 15.853 17 7.4000 4.5000 0.9000 14.296 15 2.8000 0.9000 1.8000 16.271 14 1.3207 1.2474 0.7337 299.34 334 27.900 17.100 17.100 28.379 21 1.1740 0.8071 0.0734 231.01 138 Organized By: Vivekanand Education Society's Institute Of Technology International Journal of Application or Innovation in Engineering & Management (IJAI (IJAIE AIEM) Web Site: www.ijaiem.org Email: editor@ijaiem.org ISSN 2319 - 4847 Special Issue for International Technological Conference-2014 Figure 3. Step response of the Transfer function 3C .40 with the values of GH , GI and GJ parameters obtained by the existing DPAGA and the proposed DPAGA in Trial 1 Figure 4. Step response of the Transfer function 3C .40 with the values of GH , GI and GJ parameters obtained by the existing DPAGA and the proposed DPAGA in Trial 2 Figure 5. Step response of the Transfer function 3E .40 with the values of GH , GI and GJ parameters obtained by the existing DPAGA and the proposed DPAGA in Trial 1 Figure 6. Step response of the Transfer function 3E .40 with the values of GH , GI and GJ parameters obtained by the existing DPAGA and the proposed DPAGA in Trial 2 The performance parameters obtained from the DPAGA [7] based PID and the Proposed DPAGA based PID in Trial 1 are shown in Table 4 and that for Trial 2 is shown in Table 5. The comparison of performance parameters of DPAGA [7] based PID and the Proposed DPAGA based PID in terms of settling time, time taken for convergence and number of generations required for convergence in Trial 1 and Trial 2 is shown in Table 6. While the comparison in terms of percentage improvements in settling time, time taken for convergence and number of generations required for convergence in Trial 1 and Trial 2 is shown in Table 7. Organized By: Vivekanand Education Society's Institute Of Technology International Journal of Application or Innovation in Engineering & Management (IJAI (IJAIE AIEM) Web Site: www.ijaiem.org Email: editor@ijaiem.org ISSN 2319 - 4847 Special Issue for International Technological Conference-2014 Table 4. The performance parameters obtained in Trial 1 Trial 1 Plant TFs 3 .40 38 .40 3> .40 3? .40 3@ .40 3C .40 3D .40 3E .40 _L 0.63 0 0.33 1.27 0 The Existing DPAGA[7] – PID `a `b \] 0.764 17.9 30 0.0435 6.4 30 0.644 2.36 30 0.33 0.702 30 7.333 9.83 30 Unstable 30 0.05 4 2.03 30 Unstable 30 The Proposed DPAGA[7] – PID _L `a `b \] 0.3 0.363 1.27 30 0 0.0418 0.05 30 0.29 0.457 2.3 30 0.1 0.207 0.278 30 0 0.599 5.21 30 0.1 3.6 4.93 30 0.01 2.25 0.139 30 0.09 1.26 1.84 30 Table 5. The performance parameters obtained in Trial 2 Trial 2 Plant TFs 3 .40 38 .40 3> .40 3? .40 3@ .40 3C .40 3D .40 3E .40 _L 0.6 0 0.29 0.27 0 The Existing DPAGA[7] – PID `a `b \] 0.96 17.1 1104 0.1048 2.06 14 0.457 2.31 15 0.144 0.702 16 44.605 45.4 1105 Unstable 9 0.04 3.33 1.5 57 Unstable 9 The Proposed DPAGA[7] – PID _L `a `b \] 0.3 0.363 1.27 9 0 0.0418 0.05 22 0.29 0.457 2.31 17 0.1 0.207 0.278 15 0 0.599 5.21 14 0.1 3.6 4.93 334 0.01 2.25 0.139 21 0.13 0.895 1.84 138 Table 6. The comparison of the improvement in the performance parameters obtained in Trial 1 and Trial 2 The Existing DPAGA[7] – PID Plant TFs 3 .40 38 .40 3> .40 3? .40 3@ .40 3C .40 3D .40 3E .40 Number of Generations Trial 1 Trial 2 30 1104 30 14 30 15 30 16 30 1105 30 9 30 57 30 9 Settling time Trial 1 17.9 6.4 2.36 0.702 9.83 unstable 2.03 unstable Trial 2 17.1 2.06 2.31 0.702 45.4 unstable 1.5 unstable The Proposed DPAGA[7] – PID Number of Generations Trial 1 Trial 2 30 9 30 22 30 17 30 15 30 14 30 334 30 21 30 138 Settling time Trial 1 1.27 0.05 2.31 0.278 5.21 4.93 0.139 1.84 Trial 2 1.27 0.05 2.31 0.278 5.21 4.93 0.139 1.84 The Table 7 indicates that the Proposed DPAGA gives very good improvements in the settling time. In case of 38 .40 and 3C .40 the Proposed DPAGA takes more convergence time and more number of generations compared to the existing DPAGA but, the performance of the Proposed DPAGA is better than the existing DAPGA. Also, in case of 3C .40 and 3E .40 the Proposed DPAGA takes more convergence time and more number of generations compared to the existing DPAGA but, the performance of the Proposed DPAGA is acceptable while the existing DAPGA gives unstable response in Trial 1 and Trial 2. Organized By: Vivekanand Education Society's Institute Of Technology International Journal of Application or Innovation in Engineering & Management (IJAI (IJAIE AIEM) Web Site: www.ijaiem.org Email: editor@ijaiem.org ISSN 2319 - 4847 Special Issue for International Technological Conference-2014 Table 7. The comparison of the percentage improvements in the performance parameters obtained in Trial 1 and Trial 2 Settling time Plant TFs 3 .40 38 .40 3> .40 3? .40 3@ .40 3C .40 3D .40 3E .40 Trial 1 92.905% 99.218% 2.542% 70.085% 46.998% ∞% 93.152% ∞% Trial 2 92.573% 97.572% 0% 60.398% 88.722% ∞% 90.733% ∞% Convergence time Trial 1 32.223% 32.040% 33.8277% 31.906% 31.089% 55.684% 31.169% 31.704% Trial 2 99.403% -12.511% 20.115% 64.777% 98.836% -24.922% 73.320% 1169.84% Number of Generations Trial 1 NA NA NA NA NA NA NA NA Trial 2 99.1364% -57.142% -5.686% 64.777% 98.9376% -1611.11% 63.157% -1433.33% 5. CONCLUSION In our work we tried to improve the performance of the existing DPAGA in terms of number of iterations and settling time. We suggest to check the convergence condition after evaluating the individuals of both the populations i.e. the main population and the subordinate population. This improves the performance of the existing algorithm. We also applied these two algorithms to optimize the PID parameters. The performance of the proposed DPAGA based PID controllers gave acceptable results for all the selected transfer functions. Moreover, in almost all the cases the Proposed DPAGA based PID performed better than the existing DPAGA based PID in terms of settling time, number of iterations and time taken for convergence. We observed that in few of the cases the tuning parameters obtained using existing DPAGA gave unstable response whereas for the same transfer functions the proposed algorithm gave stable response. 6. FUTURE SCOPE The scheme of DPAGA may be adopted for optimization of the PID Parameters in industrial controllers. References [1] D. E. Goldberg and J. Richardson, “Genetic algorithms with sharing for multimodal function optimization,” in Proc. 2nd Int. Conf. Genetic Algorithms (ICGA), pp. 41–49, 1987. [2] Y. Jie, N. Kharma, and P. Grogono, “BMPGA: A bi-objective multipopulation genetic algorithm for multimodal function optimization,” in Proc. IEEE Congr. Evol. Comput., vol. 1. , pp. 816–823, 2005. [3] T. Park and K. R. Ryu. “A dual population genetic algorithm with evolving diversity.” In IEEE Congress on Evolutionary Computation (CEC2007), pages 3516–3522, 2007. [4] T. Park, R. Choe, and K. R. Ryu. “Adjusting population distance for the dual-population genetic algorithm.” In Australian Conference on Artificial Intelligence (AI 2007) (LNCS 4830), pages 171–180, 2007. [5] T. Park, and K. R. Ryu, "A dual-population genetic algorithm for balanced exploration and exploitation," in Proc. of Computational Intelligence, pp. 88-93, 2006. [6] T. Park, and K. R. Ryu, “A Dual-Population Genetic Algorithm for Adaptive Diversity Control,” Transactions on Evolutionary Computation, vol. 14, pp.865 – 884, 2010. [7] YAN Tai-shan, GUO Guan-qi, LI Wu “Research on A Novel Genetic Algorithm Based on Adaptive Evolution in Dual Population,” IEEE Conference Publications, International Conference on Consumer Electronics, Communications and Networks (CECNet), pp. 594 - 597, 2011. [8] M. Wineberg and F. Oppacher, "Distances between populations," in Proc. of the Fifth Genetic and Evolutionary Computation Conference, pp. 1481-1492, 2003. [9] P. J. Van Rensberg, I. S. Shaw, J. D. van Wyk, “Adaptive PID control using a genetic algorithm,” Second International Conference on Knowledge – Based Intelligence Electronic Systems, pp. 133-138, 1998. [10] Liu Fan Er Meng Joo, “Design of Auto tuning PID controller based on Genetic Algorithm”, IEEE Conference on Industrial Electronics and Applications, ICIEA, 2009. Organized By: Vivekanand Education Society's Institute Of Technology International Journal of Application or Innovation in Engineering & Management (IJAI (IJAIE AIEM) Web Site: www.ijaiem.org Email: editor@ijaiem.org ISSN 2319 - 4847 Special Issue for International Technological Conference-2014 [11] Gouhab Lin, Guofan Liu, “Tuning PID Controller using Adaptive Genetic Algorithms” The 5th International Conference on Computer Science & Education Hefei, China, August 24-27, 2010. AUTHORS Shanta Sondur received the B.E. and M.E. degrees in Instrumentation and Control Engineering from S.G.G.S.C. of Engineering and Technology, Nanded and Electronics and Telecommunications from Govt. C.O.E Pune, in 1989 and 1997, respectively. She has received PhD degree from Systems and Control Engineering, IIT Bombay in 2008. During 1991-1998, she worked as faculty in Dept. of Instrumentation at GCOE, Pune. Now she is working as a Professor in Dept. of IT, at VES Institute of Technology, Mumbai, India. Anil Kumar R. Sudele received the B.E. degree in Electrical Engineering from Yadavrao Tasgaonkar Institute of Engineering and Technology, Karjat and M.E. degree in Instrumentation & Control Engineering from VES Institute of Technology, Mumbai in 2010 and 2013, respectively. From 2013 he is working as Core Resource Group Engineer at SMEC Automation Pvt. Ltd., Mumbai, India. His areas of interest are Process control and Automation. Organized By: Vivekanand Education Society's Institute Of Technology