Evolutionary Algorithm for Artificial-Immune-System-Based Failure-Detector Generation and Optimization

Abstract

The development of an evolutionary algorithm and accompanying software for the generation and optimization of artificial-immune-system-based failure detectors using the negative-selection strategy is presented in this paper. A detector is defined as a subregion of the hyperspace formed by relevant system parameters at abnormal conditions. The utility is a part of an integrated set of methodologies for the detection, identification, and evaluation of a wide variety of aircraft subsystem abnormal conditions. The process of generating and optimizing detectors has several phases. A preliminary phase consists of processing data from flight tests for self definition, including normalization, duplicate removal, and clustering. A first phase of the evolutionary algorithm produces, through an iterative process, a set of detectors that do not overlap with the self and achieve a prescribed level of coverage of the nonself. A second phase consists of a classic genetic algorithm that attempts to optimize the number of detectors, overlapping between detectors, and coverage of the nonself while maintaining no overlap with the self. For this second phase, an initial individual is a set of detectors obtained in the first phase. Specific genetic operators have been defined to accommodate different detector shapes, such as hyperrectangles, hyperspheres, and hyperellipsoids. An interactive design environment has been developed in MATLAB that relies on an advanced user-friendly graphical interface and on a substantial library of alternative algorithms to allow maximum flexibility and effectiveness in the design of detector sets for artificial-immune-system-based abnormal-condition detection. The desirable performance of the proposed methodology is demonstrated by comparing the detection results for aircraft actuator failures of two unoptimized detector sets with the detection results of an optimized detector set. These results show that the algorithm can determine equal or better detection performance while using fewer detectors to cover the nonself. Nomenclature a = semi-axis length vector c = center D = detector, genotypes d = semi-side length vector, applicable to rectangles L = lower, or worse, limit N = population size PI = performance-index value p = probability of selection q = cumulative probability of selection r = radius S = self S = nonself TF = total fitness U = upper, or better, limit W = performance-index weight = universe, all possible solutions Subscripts coverage = coverage performance-index criterion E = hyperellipsoids e = hyperellipsoids i = referring to a value for a particular individual number = number of detectors performance-index criterion overlap = overlap performance-index criterion S = hyperspheres s = hyperspheres R = hyperrectangles RE = hyperrotational ellipsoids r = hyperrectangles re = hyperrotational ellipsoids

Integrated Immunity-Based Framework for Aircraft Abnormal Conditions Management

“…A set of computational tools have been developed at WVU for the generation, optimization, and verification of detector sets within the AIS paradigm [32,33]. The WVU immunity-based failure detector optimization and testing tool relies primarily on evolutionary computation providing a wide selection of algorithms and options as well as capabilities for testing and tuning [26,34].…”

“…The immunitybased fault detection [22,23] operates in a similar manner as does the biological immune system, according to the principle of self/nonself discrimination, when it detects exogenous antigens while not reacting to the self cells. An integrated set of methodologies [24][25][26][27] for AIS-based detection and identification of a wide variety of aircraft subsystem failures/damages has been designed and implemented at West Virginia University (WVU). Integrated high-performance AISbased failure detection and identification schemes have been demonstrated to be capable of handling several categories of subsystem abnormal conditions over extended areas of the flight envelope [28].…”

Aircraft Abnormal Conditions Detection, Identification, and Evaluation Using Innate and Adaptive Immune Systems Interaction

“…The parameters related to the self/nonself generation (such as the shape, number, and size of the self clusters and the overlapping between self clusters and detectors) affect both the detection performance of the subselves and the computational resources required by each method. To achieve better detection performance with less computational resources, these parameters need to be tuned through optimization (for example, by using the genetic algorithm suggested in [84]). Additional nominal flight tests covering a wider range of the flight envelope are expected to improve the overall detection performance of the generated subselves but at the expense of the computer memory and computational speed.…”

“…The potential of the artificial immune system to provide adaptive control of a UAV has been recently investigated by augmenting an immunity-based mechanism to the nonlinear dynamic inversion (NLDI) of the UAV in an attempt to provide adaptive control laws [82]. An evolutionary algorithm has been developed for the generation and optimization of artificial immune system-based failure detectors using the negative selection strategy [83,84].…”

“…where D j is the center and Rd j is the radius of detector j. When generating the self and the nonself, the following optimization criteria should be considered [78,83,84]:…”