E. Lughofer∗ and C. Guardiola∗∗
Fault detection, changing operating conditions, flexibly evolving fuzzy models, residuals, local confidence regions, adaptive statistical methods, engine test benches, comparison with analytical models
The main contribution of this paper is a novel fault detection strategy, which is able to cope with changing system states at on-line measurement systems fully automatically. For doing so, an improved fault detection logic is introduced which is based on data-driven evolving fuzzy models. These are sample-wise trained from on-line measurement data from scratch, i.e., the structure and rules of the models evolve over time in order to cope (1) with high-frequented measurement recordings and (2) on-line changing operating conditions. The evolving models represent (changing) dependencies between certain system variables and are used for calculating the deviation between expected model outputs and real- measured values on new incoming data samples (→ residuals). The residuals are compared with confidence regions surrounding the evolving fuzzy models, so-called local error bars and their behaviour is analysed over time by adaptive univariate statistical methods → anomalies in the residual signals indicate faults in the system. Due to local error bars, it is possible to react very flexibly on local regions within the system variables and hence to increase the fault detection performance significantly. Evaluation results based on high-dimensional measurement data from engine test benches are demonstrated at the end of the paper, where the novel fault detection approach is compared against static analytical (fault) models.
Important Links:
Go Back
