Bayesian Methods for Aircraft Structural Health Monitoring

Aircraft structures, whether metallic or composite, are subject to service damage which requires their periodic inspection and maintenance. While taking the aircraft out of service is quite costly, the assurance of structural integrity requires such inspection – and possible repair work based on the inspection results obtained. If damage is detected from an inspection, the decision whether to repair as well as the method of repair must be made on the basis of data relevant to the specific inspection method being used, and with an uncertainty that has been characterized as accurately as possible. This is an example of an application that is amenable to a Dynamic Data Driven Application System (DDDAS) solution. That is, data can be acquired dynamically, and compared to a model of the structure such that damage can be located and a determination made as to whether it requires further inspection and possible repair. Moreover, Bayesian methods allow the characterization of uncertainty, and with the appropriate inference networks they allow conditional probabilities to be determined in terms of what is known about the structure from the model and what is measured during the inspection. The methodology under development allows the accuracy of the model as well as that of the inspection data to be taken into consideration, and uses an iterative approach to improve both the model and the inspection data. That is, inspection data can be used to determine physical constants or variables used in the model (e.g., Young’s modulus, diffusion constants, etc.), and the computational model can be used to improve inspection data (e.g., pose, noise, hysteresis, etc.).