LAUSR

Abstract In this paper, an adaptive fault diagnosis approach is proposed in order to perform the fault diagnosis of manufacturing systems. The desired (normal) behavior is represented by a set of temporal specifications while faults are considered to be the execution of specific fault behavior violating one or more of these specifications. The inference of the fault type of each fault is achieved by a diagnosis module called diagnoser. Diagnoser’s model is based on DEVS formalism and two conditions is given to verify if the diagnoser covers fully and optimally the observable events. The approach considers that only normal behavior is known initially and therefore the diagnoser will be generate firstly with just this normal behavior. Since, it is hard to include in advance all abnormal behaviors, then the approach adapts the diagnoser in order to integrate new specific fault behaviors iteratively into its inference engine. This adaptation allows increasing the diagnosis capacity, called diagnosability, over time. Real-time tests are performed on an automated system available into our lab to allow an online validation.