David W. Coit

Multi-objective optimization using genetic algorithms: A tutorial

Multi-objective formulations are realistic models for many complex engineering optimization problems. In many real-life problems, objectives under consideration conflict with each other, and optimizing a particular solution with respect to a single objective can result in unacceptable results with respect to the other objectives. A reasonable solution to a multi-objective problem is to investigate a set of solutions, each of which satisfies the objectives at an acceptable level without being dominated by any other solution. In this paper, an overview and tutorial is presented describing genetic algorithms (GA) developed specifically for problems with multiple objectives. They differ primarily from traditional GA by using specialized fitness functions and introducing methods to promote solution diversity.

Reliability optimization of series-parallel systems using a genetic algorithm

A problem-specific genetic algorithm (GA) is developed and demonstrated to analyze series-parallel systems and to determine the optimal design configuration when there are multiple component choices available for each of several k-out-of-n:G subsystems. The problem is to select components and redundancy-levels to optimize some objective function, given system-level constraints on reliability, cost, and/or weight. Previous formulations of the problem have implicit restrictions concerning the type of redundancy allowed, the number of available component choices, and whether mixing of components is allowed. GA is a robust evolutionary optimization search technique with very few restrictions concerning the type or size of the design problem. The solution approach was to solve the dual of a nonlinear optimization problem by using a dynamic penalty function. GA performs very well on two types of problems: (1) redundancy allocation originally proposed by Fyffe, Hines, Lee, and (2) randomly generated problem with more complex k-out-of-n:G configurations.

Penalty guided genetic search for reliability design optimization

Reliability optimization has been studied in the literature for decades, usually using a mathematical programming approach. Because of these solution methodologies, restrictions on the type of allowable design have been made, however heuristic optimization approaches are free of such binding restrictions. One difficulty in applying heuristic approaches to reliability design is the highly constrained nature of the problems, both in terms of number of constraints and the difficulty of satisfying constraints. This paper presents a penalty guided genetic algorithm which efficiently and effectively searches over promising feasible and infeasible regions to identify a final, feasible optimal, or near optimal, solution. The penalty function is adaptive and responds to the search history. Results obtained on 33 test problems from the literature dominate previous solution techniques.

Adaptive Penalty Methods for Genetic Optimization of Constrained Combinatorial Problems

The application of genetic algorithms (GA) to constrained optimization problems has been hindered by the inefficiencies of reproduction and mutation when feasibility of generated solutions is impossible to guarantee and feasible solutions are very difficult to find. Although several authors have suggested the use of both static and dynamic penalty functions for genetic search, this paper presents a general adaptive penalty technique which makes use of feedback obtained during the search along with a dynamic distance metric. The effectiveness of this method is illustrated on two diverse combinatorial applications: (1) the unequal-area, shape-constrained facility layout problem and (2) the series-parallel redundancy allocation problem to maximize system reliability given cost and weight constraints. The adaptive penalty function is shown to be robust with regard to random number seed, parameter settings, number and degree of constraints, and problem instance.

A Monte-Carlo simulation approach for approximating multi-state two-terminal reliability

This paper describes a Monte-Carlo (MC) simulation methodology for estimating the reliability of a multi-state network. The problem under consideration involves multi-state two-terminal reliability (M2TR) computation. Previous approaches have relied on enumeration or on the computation of multi-state minimal cut vectors (MMCV) and the application of inclusion/exclusion formulae. This paper discusses issues related to the reliability calculation process based on MMCV. For large systems with even a relatively small number of component states, reliability computation can become prohibitive or inaccurate using current methods. The major focus of this paper is to present and compare a new MC simulation approach that obtains accurate approximations to the actual M2TR. The methodology uses MC to generate system state vectors. Once a vector is obtained, it is compared to the set of MMCV to determine whether the capacity of the vector satisfies the required demand. Examples are used to illustrate and validate the methodology. The estimates of the simulation approach are compared to exact and approximation procedures from solution quality and computational effort perspectives. Results obtained from the simulation approach show that for relatively large networks, the maximum absolute relative error between the simulation and the actual M2TR is less than 0.9%, yet when considering approximation formulae, this error can be as large as 18.97%. Finally, the paper discusses that the MC approach consistently yields accurate results while the accuracy of the bounding methodologies can be dependant on components that have considerable impact on the system design.

Cold-standby redundancy optimization for nonrepairable systems

Abstract A solution methodology is described and demonstrated to determine optimal design configurations for nonrepairable series-parallel systems with cold-standby redundancy. This problem formulation considers non-constant component hazard functions and imperfect switching. The objective of the redundancy allocation problem is to select from available components and to determine an optimal design configuration to maximize system reliability. For cold-standby redundancy, other formulations have generally required exponential component time-to-failure and perfect switching assumptions. For this paper, there are multiple component choices available for each subsystem and component time-to-failure is distributed according to an Erlang distribution. Optimal solutions are determined based on an equivalent problem formulation and integer programming. Compared to other available algorithms, the methodology presented here more accurately models many engineering design problems with cold-standby redundancy. Previously, it has been difficult to determine optimal solutions for this class of problems or even lo efficiently calculate system reliability. The methodology is successfully demonstrated on a large problem with 14 subsystems.

Reliability and maintenance modeling for systems subjected to multiple dependent competing failure precesses

For complex systems that experience Multiple Dependent Competing Failure Processes (MDCFP), the dependency among the failure processes presents challenging issues in reliability modeling. This article, develops reliability models and preventive maintenance policies for systems subject to MDCFP. Specifically, two dependent/correlated failure processes are considered: soft failures caused jointly by continuous smooth degradation and additional abrupt degradation damage due to a shock process and catastrophic failures caused by an abrupt and sudden stress from the same shock process. A general reliability model is developed based on degradation and random shock modeling (i.e., extreme and cumulative shock models), which is then extended to a specific model for a linear degradation path and normally distributed shock load sizes and damage sizes. A preventive maintenance policy using periodic inspection is also developed by minimizing the average long-run maintenance cost rate. The developed reliability and maintenance models are demonstrated for a micro-electro-mechanical systems application example. These models can also be applied directly or customized for other complex systems that experience multiple dependent competing failure processes.

Maximization of System Reliability with a Choice of Redundancy Strategies

Optimal solutions to the redundancy allocation problem are determined when either active or cold-standby redundancy can be selectively chosen for individual subsystems. This problem involves the selection of components and redundancy levels to maximize system reliability. Previously, solutions to the problem could only be found if analysts were restricted to a predetermined redundancy strategy for the complete system. Generally, it had been assumed that active redundancy was to be used. However, in practice both active and cold-standby redundancy may be used within a particular system design and the choice of redundancy strategy becomes an additional decision variable. Available optimization algorithms are inadequate for these design problems and better alternatives are required. The methodology presented here is specifically developed to accommodate the case where there is a choice of redundancy strategy. The problem is formulated with imperfect sensing and switching of cold-standby redundant components and k -Erlang distributed time-to-failure. Optimal solutions to the problem are found by an equivalent problem formulation and integer programming. The methodology is demonstrated on a well-known test problem with interesting results. The optimal system design is distinctly different from the corresponding design obtained with only active redundancy. The availability of this tool can result in more reliable and cost-effective engineering designs.

A HEURISTIC FOR SOLVING THE REDUNDANCY ALLOCATION PROBLEM FOR MULTI-STATE SERIES–PARALLEL SYSTEMS

Abstract The redundancy allocation problem is formulated with the objective of minimizing design cost, when the system exhibits a multi-state reliability behavior, given system-level performance constraints. When the multi-state nature of the system is considered, traditional solution methodologies are no longer valid. This study considers a multi-state series-parallel system (MSPS) with capacitated binary components that can provide different multi-state system performance levels. The different demand levels, which must be supplied during the system-operating period, result in the multi-state nature of the system. The new solution methodology offers several distinct benefits compared to traditional formulations of the MSPS redundancy allocation problem. For some systems, recognizing that different component versions yield different system performance is critical so that the overall system reliability estimation and associated design models the true system reliability behavior more realistically. The MSPS design problem, solved in this study, has been previously analyzed using genetic algorithms (GAs) and the universal generating function. The specific problem being addressed is one where there are multiple component choices, but once a component selection is made, only the same component type can be used to provide redundancy. This is the first time that the MSPS design problem has been addressed without using GAs. The heuristic offers more efficient and straightforward analyses. Solutions to three different problem types are obtained illustrating the simplicity and ease of application of the heuristic without compromising the intended optimization needs.

MOMS-GA: A Multi-Objective Multi-State Genetic Algorithm for System Reliability Optimization Design Problems

A custom genetic algorithm was developed and implemented to solve multiple objective multi-state reliability optimization design problems. Many real-world engineering design problems are multi-objective in nature, and among those, several of them have various levels of system performance ranging from perfectly functioning to completely failed. This multi-objective genetic algorithm uses the universal moment generating function approach to evaluate the different reliability or availability indices of the system. The components are characterized by having different performance levels, cost, weight, and reliability. The solution to the multi-objective multi-state problem is a set of solutions, known as the Pareto-front, from which the analyst may choose one solution for system implementation. Two illustrative examples are presented to show the performance of the algorithm; and the multi-objective formulation considered for both of them is the maximization of system availability, and the minimization of both system cost, and weight.