Garrison W. Greenwood

Searching for multiobjective preventive maintenance schedules: Combining preferences with evolutionary algorithms

Abstract Heavy industry maintenance facilities at aircraft service centers or railroad yards must contend with scheduling preventive maintenance tasks to ensure critical equipment remains available. The workforce that performs these tasks are often high-paid, which means the task scheduling should minimize worker idle time. Idle time can always be minimized by reducing the workforce. However, all preventive maintenance tasks should be completed as quickly as possible to make equipment available. This means the completion time should be also minimized. Unfortunately, a small workforce cannot complete many maintenance tasks per hour. Hence, there is a tradeoff: should the workforce be small to reduce idle time or should it be large so more maintenance can be performed each hour? A cost effective schedule should strike some balance between a minimum schedule and a minimum size workforce. This paper uses evolutionary algorithms to solve this multiobjective problem. However, rather than conducting a conventional dominance-based Pareto search, we introduce a form of utility theory to find Pareto optimal solutions. The advantage of this method is the user can target specific subsets of the Pareto front by merely ranking a small set of initial solutions. A large example problem is used to demonstrate our method.

Workforce‐constrained Preventive Maintenance Scheduling Using Evolution Strategies

Heavy equipment overhaul facilities such as aircraft service centers and railroad yards face the challenge of minimizing the makespan for a set of preventive maintenance (PM) tasks, requiring single or multiple skills, within workforce availability constraints. In this paper, we examine the utility of evolution strategies to this problem. Comparison of the computational efforts of evolution strategies with exhaustive enumeration to reach optimal solutions for 60 small problems illustrates the ability of evolution strategies to yield optimal solutions increasingly efficiently with increasing problem size. A set of 852 large-scale problems was solved using evolution strategies to examine the effects of task-related problem characteristics, workforce-related variables, and evolution strategies population size (μ) on CPU time. The results empirically supported practical utility of evolution strategies to solve large-scale, complex preventive maintenance problems involving single- and multiple-skilled workforce. Finally, comparison of evolution strategies and simulated annealing for the 852 experiments indicated much faster convergence to optimality with evolution strategies.

Preference-driven hierarchical hardware/software partitioning

We present a hierarchical evolutionary approach to hardware/software partitioning for real-time embedded systems. In contrast to most previous approaches, we apply a hierarchical structure and dynamically determine the granularity of tasks and hardware modules to adaptively optimize the solution while keeping the search space as small as possible. Two new search operators are described, which exploit the proposed hierarchical structure. Efficient ranking is another problem addressed. Imprecisely specified multiple attribute utility theory has the advantage of constraining the solution space computation overhead. We propose a new technique to reduce the overhead. Experiment results show that our algorithm is both effective and efficient.

Convergence in Evolutionary Programs with Self-Adaptation

Evolutionary programs are capable of finding good solutions to difficult optimization problems. Previous analysis of their convergence properties has normally assumed the strategy parameters are kept constant, although in practice these parameters are dynamically altered. In this paper, we propose a modified version of the 1/5-success rule for self-adaptation in evolution strategies (ES). Formal proofs of the long-term behavior produced by our self-adaptation method are included. Both elitist and non-elitist ES variants are analyzed. Preliminary tests indicate an ES with our modified self-adaptation method compares favorably to both a non-adapted ES and a 1/5-success rule adapted ES.

Training partially recurrent neural networks using evolutionary strategies

This correspondence presents the latest results of using evolutionary strategies (ESs) to design partially recurrent neural networks for viseme recognition. ESs are stochastic optimization algorithms based upon the principles of natural selection found in the biological world. Our results indicate that ESs can be effectively used to determine the synaptic weights in neural networks and can outperform backpropagation techniques.

On the practicality of using intrinsic reconfiguration for fault recovery

Evolvable hardware (EHW) combines the powerful search capability of evolutionary algorithms with the flexibility of reprogrammable devices, thereby providing a natural framework for reconfiguration. This framework has generated an interest in using EHW for fault-tolerant systems because reconfiguration can effectively deal with hardware faults whenever it is impossible to provide spares. But systems cannot tolerate faults indefinitely, which means reconfiguration does have a deadline. The focus of previous EHW research relating to fault-tolerance has been primarily restricted to restoring functionality, with no real consideration of time constraints. In this paper, we are concerned with EHW performing reconfiguration under deadline constraints. In particular, we investigate reconfigurable hardware that undergoes intrinsic evolution. We show that fault recovery done by intrinsic reconfiguration has some restrictions, which designers cannot ignore.

Using Differential Evolution for a Subclass of Graph Theory Problems

Conventional differential evolution algorithms explore continuous spaces. In contrast, NP-complete graph problems are combinatorial and thus have discrete solution spaces, many with solutions encoded as binary strings. This letter describes a simple method showing how a conventional differential evolution algorithm can be used to solve these types of graph theory problems. The method is deterministic and can handle graphs of arbitrary size.

Evolutionary computation with extinction: experiments and analysis

Under a species-level abstraction of classical evolutionary programming, the standard tournament selection model is not appropriate. When viewed in this manner, it is more appropriate to consider two modes of life histories: background evolution and extinction. The utility of this approach as an optimization procedure is evaluated on a series of test functions relative to the performance of classical evolutionary programming and fast evolutionary programming. The results indicate that on some smooth, convex landscapes and over noisy, highly multimodal landscapes, extinction evolutionary programming can outperform classical and fast evolutionary programming. On other landscapes, however, extinction evolutionary programming performs considerably worse than classical and fast evolutionary programming. Potential reasons for this variability in performance are indicated.

Scheduling tasks in real-time systems using evolutionary strategies

Finding feasible schedules for tasks running in hard, real-time distributed computing systems is generally NP-hard. This paper describes a heuristic algorithm using evolutionary strategies. Our results indicate that the evolutionary strategies can find feasible schedules (assuming they exist) in very short periods of time.<<ETX>>

Scheduling for power reduction in a real-time system

This paper describes how, through a combination of scheduling and buffer insertion, real-time systems may be optimized for power consumption while maintaining deadlines. Beginning with simple examples (components that have no internal pipelines and in which the only design freedoms are buffer insertion and scheduling), we illustrate the effect of adjusting the time at which data are processed on power consumption. Algorithms for optimizing the energy saving are proposed for several real-time system implementations including non-pipelined and pipelined. We also discuss extension to this preliminary work including selection of alternate processing units in order to reduce power consumption while maintaining deadlines.