D. Powell and M. Skolnick. Using genetic algorithms in engineering design optimization with non-linear constraints. In Proceedings of the Fifth International Conference on Genetic Algorithms, pages 424--431. Morgan Kaufmann, July 1993.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....individual satisfying the jth constraint we can always count with all individuals of the previous generation. This insures that the size of the population remains constant even when eliminating those individuals which violate the constraints. 2. 4 Method P This method was developed circa 1993 [6] and includes a heuristic to single out nonfeasible points: Any feasible solution is better than a non feasible one . The penalty function is defined as follows: 0 ) 1 j j M x M x p j j x f x f min x f max max t x M x t x x f r ....

Powell, D. and Skolnick, M., "Using Genetic Algorithms in Engineering Design Optimization with Non-linear Constraints". Proceedings of the Fifth International Conference on Genetic Algorithms. pp. 424-430, 1993.

....them) are equally dicult to satisfy. Also, there is no further evidence of the e ectiveness of the approach in 18 other combinatorial optimization problems, and apparently, it has not been extended to numerical optimization problems either. 5. 2 Superiority of feasible points Powell and Skolnick [134] incorporated a heuristic rule (suggested by Richardson et al. 144] for processing infeasible solutions: evaluations of feasible solutions are mapped into the interval (1, 1) and infeasible solutions into the interval (1, 1) Individuals are evaluated using [134] f( x) if feasible 1 r P ....

.... points Powell and Skolnick [134] incorporated a heuristic rule (suggested by Richardson et al. 144] for processing infeasible solutions: evaluations of feasible solutions are mapped into the interval (1, 1) and infeasible solutions into the interval (1, 1) Individuals are evaluated using [134]: f( x) if feasible 1 r P n i=1 g i ( x) P p j=1 h j ( x) 39) f( x) is scaled into the interval (1,1) g i ( x) and h j ( x) are scaled into the interval (1, 1) and r is a constant. Notice that in this approach the objective function and the amount of constraint violation are not ....

[Article contains additional citation context not shown here]

David Powell and Michael M. Skolnick. Using genetic algorithms in engineering design optimization with nonlinear constraints. In Stephanie Forrest, editor, Proceedings of the Fifth International Conference on Genetic Algorithms, pages 424-431, San Mateo, California, July 1993. University of Illinois at Urbana-Champaign, Morgan Kaufmann Publishers.

....; f(X ) After some number of generations the temperature is decreased and the best solution found so far serves as a starting point of the next iteration. The process continues until the temperature reaches the freezing point. The next method was developed by Powell and Skolnick [15]. The method is a classical penalty method with one notable exception. Each individual is evaluated by the formula: eval(X ) f(X ) r j=1 f j (X) t; X) where r is a constant; however, there is also a component (t; X) This is an additional iteration dependent function which influences ....

.... of results (between 680.642 and 689.660, see [10] Of course, all resulting points X were feasible, which was not the case with other systems (e.g. Genocop II produced a value of 18.917 for the problem G5, the systems based on the methods of Homaifar, Lai, and Qi [6] and Powell and Skolnick [15] gave results of 2282.723 and 2101.367, respectively, for the problem G2) Clearly, Genocop III is a promising tool for constrained nonlinear optimization problems. However, there are many issues which require further attention and experiments. These include investigation of the significance of ....

Powell, D. and Skolnick, M.M., Using Genetic Algorithms in Engineering Design Optimization with Non-linear Constraints, Proceedings of the Fifth ICGA, Morgan Kaufmann, 1993, pp.424--430.

....reported to handle general nonlinear programming problems. However, their description is supported by experimental evidence based on different test cases; some of them provide results for test cases with very few variables [11, 16, 7] some of them did not use functions which have a closed form [14]. Also, they differ in details (representation, selection method, operators and their frequencies, etc. so it is quite hard to make any comparisons. This paper surveys these methods (next Section) and provides five test cases (Section 3) which are used here to test the methods listed, and can ....

....point. The method requires a starting and freezing temperatures, 0 and f , respectively, and the cooling scheme to decrease temperature . Standard values (reported in [11] are 0 = 1, i 1 = 0:1 Delta i , with f = 0:000001. The fifth method was developed by Powell and Skolnick [14]. The method is a classical penalty method with one notable exception. Each individual is evaluated by the formula: This feature, however, is not essential. The only important requirement is that the next population contains the best individual from the previous population. eval(X ) f(X ) ....

Powell, D. and Skolnick, M.M., Using Genetic Algorithms in Engineering Design Optimization with Non-linear Constraints, Proceedings of the Fifth ICGA, Morgan Kaufmann, 1993, pp.424-- 430.

....INSA CNRS UMR 6138, Rouen, France. E mail: Rodolphe.Leriche insa rouen.fr . F. Guyon is with the Bio statistics Bio mathematics lab. Paris 7 univ. France. E mail: guyon urbb.jussieu.fr . Four types of methods for handling constraints exist: penalization of infeasible solutions( 8] 11] [23], 25] 29] 2] 6] projection of infeasible solutions onto the feasible domain ( 20] 27] 28] co evolution of populations which together solve the constrained optimization problem ( 22] and constraints representation building in the course of the search ( 21] 24] These approaches ....

....strategies, one distinguishes static, dynamic and adaptive methods. Static penalties depend neither on the number of points sampled during the search nor on their performance ( 8] 13] Dynamic penalties ( 16] 10] are function of the number of points sampled while adaptive penalties ( 2] 6] [23], 11] vary with points evaluations. Mixed approaches exist, e.g. in [29] Duality and related concepts such as Lagrange multipliers have yielded some of the most ecient general purpose methods of mathematical programming for continuous, di erentiable and locally convex problems ( 19] 15] 26] ....

D. Powell and M.M. Skolnick, \Using genetic algorithms in engineering design optimization with non-linear constraints", Proc. of the Fourth International Conference on Genetic Algorithms, San Mateo, CA, Morgan Kaufmann, 1991, pp. 424-431.

....related is schedule optimization, or time tabling, where the task is to allocate efficiently a set of resources to carry out a set of tasks [5, 6, 14] design optimization Design tasks can be seen as a mix of combinatorial and function optimization. Design applications include engineering design [22], communication network design [3, 4] and neural network design [10] machine learning There are many applications of GAs to learning systems, the usual paradigm being that of a classifier system [13] In such systems the learning process is controlled by a GA that tries to evolve a population ....

....to avoid the risk of exceeding maximum inventory levels. These constraints are handled by penalizing unfeasible individuals under the assumption of the superiority of feasible solutions over unfeasible solutions [19] This is also what is done in the method described by Powell and Skolnick in [22] (suggested earlier by Richardson et al. in [23] This constraint handling technique is implemented in the following way: evaluations of feasible solutions are mapped into the interval [0; C max ] and unfeasible solutions into the interval (C max ; C max P max ] where P max is the maximum ....

David Powell and Michael M. Skolnick. Using genetic algorithms in engineering design optimization with non-linear constraints. In S. Forrest, editor, Proceedings of the Fifth International Conference on Genetic Algorithms, pages 424--431. Morgan Kaufmann, 1993.

....nonlinear constraints, feasible regions may be disconnected, and keeping a search within a feasible region may lead to poor solutions. In addition, it is very di#cult or expensive to project a trajectory into feasible regions for nonlinear constraints. Global Search. Rejecting discarding methods [110, 14, 160, 154] are stochastic procedures. They iteratively generate random points and only accept feasible points, while dropping infeasible points during their search. Although they are simple and easy to implement, they are very ine#cient when constraints are nonlinear and feasible regions are di#cult to ....

....when the objective and constraint functions are di#erentiable. Figure 2.3 classifies existing search methods for solving discrete constrained NLPs. 2.2.1 Direct Solutions for Discrete NLPs Global Search. Direct solution methods try to directly solve (1. 1) based either on rejecting discarding [110, 14, 160, 154] infeasible points or on repairing [113, 142] infeasible points into feasible ones. Both methods are not e#cient in handling nonlinear constraints, because the former wastes a lot of time in generating and rejecting infeasible points whereas the latter is very problem specific and has high ....

[Article contains additional citation context not shown here]

D. Powell and M. M. Skolnick. Using genetic algorithms in engineering design optimization with non-linear constraints. In Proc. of 5th Int'l Conf. on Genetic Algorithms, pages 424--431, 1993.

....CLM dn . 2.1.2 Direct methods for solving discrete constrained NLPs Direct solution methods for solving discrete constrained NLPs without any transformation on their objective and constraint functions can be classified into two approaches. One major approach is based on rejecting, discarding [32, 33, 126] or repairing [98] methods that try to avoid infeasible points. This approach, however, has di#culty in handling nonlinear 16 constraints whose feasible regions may be very hard to locate, leading to the generation of mostly infeasible points that are rejected. The other approach is based on ....

D. Powell and M. M. Skolnick. Using genetic algorithms in engineering design optimization with non-linear constraints. In Proc. of 5th Int'l Conf. on Genetic Algorithms, pages 424--431, 1993.

....to find. 2.1.3 Direct Solutions for Solving Discrete Constrained NLPs Direct solution methods for solving discrete constrained NLPs without any transformation on their objective and constraint functions can be classified into two approaches. One major approach is based on rejecting, discarding [8, 9, 150] or repairing [114] methods in order to avoid infeasible points. This approach, however, has di#culty in handling nonlinear constraints whose feasible regions may be very hard to locate, leading to mostly infeasible points generated that are rejected. The other approach is based on enumeration or ....

....Note that although not very practical, each algorithmic step of these methods does guide a search to achieve the goal of finding feasible solutions. Global Search methods introduce techniques to overcome local minima. Typical global search methods include rejecting methods, discarding methods [156, 150], repair methods [114, 143] and preserving feasibility [134, 76] Rejecting and discarding methods have been discussed in Section 2.1.3. Typical repair methods have some techniques to transform or repair infeasible points into feasible ones. These techniques, however, are quite limited and have ....

[Article contains additional citation context not shown here]

D. Powell and M. M. Skolnick. Using genetic algorithms in engineering design optimization with non-linear constraints. In Proc. of 5th Int'l Conf. on Genetic Algorithms, pages 424--431, 1993.

....15 2.1.1 Direct methods for solving discrete constrained NLPs Direct solution methods for solving discrete constrained NLPs without any transformation on their objective and constraint functions can be classified into two categories. One major method is based on rejecting, discarding [27, 28, 126] or repairing [97] in order to avoid infeasible points. Constraint relaxation is especially helpful for this kind of approach because after relaxation, the feasible region is enlarged, and it is much easier to find a feasible points. The other approach is based on enumeration or randomized search ....

....search to start from a feasible point. Each search step will remain in a feasible region, while trying to improve the objective function at the same time. Global Search methods use techniques to escape from local minima. Typical global search methods include rejecting methods, discarding methods [131, 126], repair methods [97, 120] and preserving feasibility [110, 73] However, these techniques are often of limited use and have di#culties in handling nonlinear constraints. Global Optimization methods use either deterministic techniques, like interval methods, or stochastic techniques, like ....

[Article contains additional citation context not shown here]

D. Powell and M. M. Skolnick. Using genetic algorithms in engineering design optimization with non-linear constraints. In Proc. of 5th Int'l Conf. on Genetic Algorithms, pages 424--431, 1993.

....done by examining the tness of feasible and infeasible members in the current population [12] The death penalty method just rejects infeasible individuals. In this method the initial population must be feasible. The method of superiority of feasible points was developed by Powell and Skolnick [13] and is based on a classical penalty approach with one notable exception. Each individual is evaluated by not only the objective and penalty function but also by an additional iteration dependent function that in uences the evaluation of infeasible solutions. The point being that, the method ....

D. Powell and M. Skolnick. Using genetic algorithms in engineering design optimization with nonlinear constraints. In S. Forrest, editor, Proc. of the 5th Int. Conf. on Genetic Algorithms, pages 424-0430. San Mateo, CA: Morgan Kaufmann, 1993.

....15 2.1.1 Direct methods for solving discrete constrained NLPs Direct solution methods for solving discrete constrained NLPs without any transformation on their objective and constraint functions can be classified into two categories. One major method is based on rejecting, discarding [27, 28, 126] or repairing [97] in order to avoid infeasible points. Constraint relaxation is especially helpful for this kind of approach because after relaxation, the feasible region is enlarged, and it is much easier to find a feasible points. The other approach is based on enumeration or randomized search ....

....a search to start from a feasible point. Each search step will remain in a feasible region, while trying to improve the objective function at the same time. Global Search methods use techniques to escape from local minima. Typical global search methods include rejecting methods, discarding methods [131, 126], repair methods [97, 120] and preserving feasibility [110, 73] However, these techniques are often of limited use and have di#culties in handling nonlinear constraints. Global Optimization methods use either deterministic techniques, like interval methods, or stochastic techniques, like ....

[Article contains additional citation context not shown here]

D. Powell and M. M. Skolnick. Using genetic algorithms in engineering design optimization with non-linear constraints. In Proc. of 5th Int'l Conf. on Genetic Algorithms, pages 424--431, 1993.

....CLM dn . 17 2.1.2 Direct methods for solving discrete constrained NLPs Direct solution methods for solving discrete constrained NLPs without any transformation on their objective and constraint functions can be classified into two approaches. One major approach is based on rejecting, discarding [32, 33, 126] or repairing [98] methods that try to avoid infeasible points. This approach, however, has difficulty in handling nonlinear constraints whose feasible regions may be very hard to locate, leading to the generation of mostly infeasible points that are rejected. The other approach is based on ....

D. Powell and M. M. Skolnick. Using genetic algorithms in engineering design optimization with non-linear constraints. In Proc. of 5th Int'l Conf. on Genetic Algorithms, pages 424--431, 1993.

....dimensionality and multi modality of the search space make it very difficult to use gradient based methods with a high degree of reliability. We also demonstrate the utility of guided crossover in a set of benchmark domains that have been used in several engineering design optimization studies [Powell and Skolnick 1993] . The remainder of this paper first presents a more detailed description of the new operator. We then present a number of experiments concerning the use of our operator on several realistic engineering design tasks. We conclude the paper with a discussion of related efforts and future work. 2 ....

....by using guided crossover. 4.3 Benchmark domains These problems were first introduced by Eric Sandgren in his Ph.D. thesis [Sandgren 1977] and have since been used in engineering design optimization research as benchmarks. One of the recent experiments involving these domains was reported in [Powell and Skolnick 1993] , in which a GA package called OOGA and a numerical optimization package called NumOpt were compared to each other in 10 of Sandgren s domains. The 10 domains were a representative sample of the original 30. We ran experiments in eight of these 10 domains. 2 All eight were minimization ....

D. Powell and M. Skolnick. Using genetic algorithms in engineering design optimization with non-linear constraints. In Proceedings of the Fifth International Conference on Genetic Algorithms, pages 424-- 431. Morgan Kaufmann, July 1993.

....performed reasonably well in all the test cases. He noted, however, that a group of methods (which he called the generalized reduced gradient methods) were reliable across a large number of test cases. The 30 problems he used were also used by other researchers before and after Sandgren s work [ Powell and Skolnick 1993 ] Those problems have now become used in engineering design optimization domains as benchmarks. One of the most recent experiments involving these domains was reported in [ Powell and Skolnick 1993 ] in which a GA package called OOGA and a numerical optimization package called NumOpt were ....

....The 30 problems he used were also used by other researchers before and after Sandgren s work [ Powell and Skolnick 1993 ] Those problems have now become used in engineering design optimization domains as benchmarks. One of the most recent experiments involving these domains was reported in [ Powell and Skolnick 1993 ] in which a GA package called OOGA and a numerical optimization package called NumOpt were compared to each other in 10 of Sandgren s domains. The 10 domains were a representative sample of the original 30. We ran experiments in eight of these 10 domains. 4 All eight were minimization ....

D. Powell and M. Skolnick. Using genetic algorithms in engineering design optimization with non-linear constraints. In Proceedings of the Fifth International Conference on Genetic Algorithms, pages 424--431. Morgan Kaufmann, July 1993.

....are compared, the one with smaller constraint violation is chosen. This approach is only applicable to population based search methods such as GAs or other evolutionary computation methods. Although at least one other constraint handling method satisfying above three criteria was suggested earlier [8] it involved penalty parameters which again must be set right for proper working of the algorithm. In the remainder of the paper, we first show that the performance of a binary coded GA using the static penalty function method on an engineering design problem largely depends on the chosen penalty ....

....is preferred to any infeasible solution, 2) Among two feasible solutions, the one having better objective function value is preferred, 7 (3) Among two infeasible solutions, the one having smaller constraint violation is preferred. Although there exist a number of other implementations [6,8,12] where criteria similar to the above are imposed in their constraint handling approaches, all of these implementations used different measures of constraint violations which still needed a penalty parameter for each constraint. Recall that penalty parameters are needed to make the constraint ....

[Article contains additional citation context not shown here]

D. Powell and M. M. Skolnick, Using genetic algorithms in engineering design optimization with nonlinear constraints. in: S. Forrest, ed., Proceedings of the Fifth International Conference on Genetic Algorithms (Morgan Kauffman, San Mateo, 1993) 424--430.

....Evolutionary algorithms can be made ecient because they are exible, and relatively easy to hybridize with domain dependent heuristics. Those features of evolutionary computation have already been acknowledged in the eld of engineering, and many applications have been reported (see, for example, [6, 11, 34, 43]) A vast majority of engineering optimization problems are constrained problems. The presence of constraints signi cantly a ects the performance of any optimization algorithm, including evolutionary search methods [20] This paper focuses on the issue of evaluation of constraints handling ....

....in the search space S. One method, proposed in [41] called a behavioral memory approach) considers the problem constraints in a sequence; a switch from one constraint to another is made upon arrival of a sucient number of feasible individuals in the population. The second method, developed in [34] is based on a classical penalty approach with one notable exception. Each individual is evaluated by the formula: eval( x) f( x) r P m j=1 f j ( x) t; x) where r is a constant; however, the original component (t; x) is an additional iteration dependent function which in uences the ....

Powell, D. and M. M. Skolnick (1993). Using genetic algorithms in engineering design optimization with non-linear constraints. In S. Forrest (Ed.), Proceedings of the 5 th International Conference on Genetic Algorithms, pp. 424-430. Morgan Kaufmann.

.... the restriction set there are several methods which can be grouped in three major categories: i) methods which preserve the feasibility of solutions [10] ii) methods based in penalty functions [16] 22] 6] 8] 12] 11] 17] and (iii) methods based in the search of feasible solutions[21][15]. Among these the method proposed in [10] is the only one with significant results when applied to high order problems. Studies conducted in [18] showed that the other two categories are perfectly suitable only when applied either to low order problems or to spaces defined by few ....

Powell, D., and Skolnick, M. Using genetic algorithms in engineering design optimization with non-linear constraints. In Proceedings of the 5th International Conference on Genetic Algorithms (San Mateo, CA, USA, July 1993), S. Forrest, Ed., Morgan Kaufmann, pp. 424--431.

....a set of special operators maintaining feasibility of solutions [9] 2.3 Adaptive methods In these methods, information is gleaned from the population to update the value of the penalty functions. A rst example of adaptive method is the method based on the superiority of feasible points [11]: Some positive term, depending on the current infeasible individuals, is added to the constraint violations in the penalty function, ensuring any feasible individual will be better than any infeasible individual. Di erent adaptive penalty methods have been proposed. In [2] the penalty coecient ....

....selection To further enhance the chances of survival of feasible individuals, a speci c selection operator is used in the algorithm. This selection, called segregational selection, can be viewed as intermediate between the method based on the superiority of feasible points (see section 2. 3 and [11]) and the replacement method used in the Segregated GA [7] see section 2.4) The segregational selection is a deterministic replacement mechanism used in an ES like scheme [14] from a population of parents, o spring are generated (all parents giving birth to o spring on average) ....

D. Powell and M. M. Skolnick. Using genetic algorithms in engineering design optimization with non-linear constraints. In S. Forrest, editor, Proc of the 5 th Int Conf on Genetic Algorithms, pages 424-430. Morgan Kaufmann, 1993.

....infeasible solutions. In fact, since finding a feasible solution to the SPP is difficult, it may be that the majority of the solutions generated by the GA are infeasible. There are two approaches to defining the fitness of an individual. The most common approach is the use of a penalty function [14, 16, 18]. Penalty methods allow constraints to be violated. Depending on the magnitude of the violation, however, a penalty that is proportional to the size of the infeasibility is incurred that degrades the objective function. If the penalty is large enough, highly infeasible individuals 3 will rarely ....

D. Powell and M. Skolnick. Using genetic algorithms in engineering design optimisation with nonlinear constraints. In S. Forrest, editor, Proceeding of the Fifth International Conference on Genetic Algorithms, pages 424--431. Morgan Kaufmann, 1993.

....in which he applied 35 nonlinear optimization algorithms to 30 engineering design optimization problems and compared their performance. Those problems have become used in engineering design optimization domains as benchmarks. One of the recent experiments involving these domains was reported in [Powell and Skolnick, 1993] , in which a GA package called OOGA and a numerical optimization package called NumOpt were compared to each other in 10 of Sandgren s do Table 1: Description of benchmark domains Domain Sandgren Dim. Constraints best No. No. inequ. equ. f 1 21 13 13 0 97.5 2 22 16 19 0 174.7 100 105 110 ....

D. Powell and M. Skolnick. Using genetic algorithms in engineering design optimization with non-linear constraints. In Proceedings of the Fifth International Conference on Genetic Algorithms, pages 424--431. Morgan Kaufmann, July 1993.

....fact that invalid individuals represent worse alternatives than valid individuals. Additive penalty functions, often used with GAs, are known to be very problem dependent and, thus, di#cult to set (Richardson et al. 1989) More recently, a more general approach based on ranking has been proposed (Powell and Skolnick, 1993). Multiple objectives Finally, control engineering problems very seldom require the optimization of a single objective function. Instead, there are often competing objectives which should be optimized simultaneously. The potential of GAs to become a powerful method for multiobjective optimization ....

....the population evolve towards a false optimum. In response to these di#culties, guidelines on the use of penalty functions have been described by Richardson et al. 1989) One of the most recent approaches to constraint handling has been proposed CHAPTER 4. MULTIOBJECTIVE GENETIC ALGORITHMS 60 by Powell and Skolnick (1993) and consists of rescaling the original objective function to assume values less than unity in the feasible region, whilst assigning infeasible individuals penalty values greater than one. Subsequently ranking the population correctly assigns higher fitness to all feasible points than to those ....

Powell, D. and Skolnick, M. M. (1993). Using genetic algorithms in engineering design optimization with non-linear constraints. In (Forrest, 1993), pages 424--431.

No context found.

D. Powell and M. Skolnick. Using genetic algorithms in engineering design optimization with non-linear constraints. In Proceedings of the Fifth International Conference on Genetic Algorithms, pages 424--431. Morgan Kaufmann, July 1993.

No context found.

Powell, David and Skolnick, M., "Using Genetic Algorithms in Engineering Design Optimization with Non-Linear Constraints," Proceedings of the Fifth International Conference on Genetic Algorithms, Morgan Kaufmann, San Mateo, CA, (1993), 426-431.

No context found.

Powell, D. and M. M. Skolnick (1993). Using genetic algorithms in engineering design optimization with non-linear constraints. In S. Forrest (Eds.), Proceedings of the 5 th International Conference on Genetic Algorithms, pp. 424--430. Morgan Kaufmann.

First 50 documents  Next 50

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC