Fault Diagnosis Based On Causal Computations

in IEEE Transactions on Systems, Man and Cybernetics. Part A, Systems and Humans (2011), 42(2), 371-381

This paper focuses on residual generation for model-based fault diagnosis. Specifically, a methodology to derive residual generators when nonlinear equations are present in the model is developed. A main ... [more ▼]

This paper focuses on residual generation for model-based fault diagnosis. Specifically, a methodology to derive residual generators when nonlinear equations are present in the model is developed. A main result is the characterization of computation sequences that are particularly easy to implement as residual generators and that take causal information into account. An efficient algorithm, based on the model structure only, which finds all such computation sequences, is derived. Furthermore, fault detectability and isolability performances depend on the sensor configuration. Therefore, another contribution is an algorithm, also based on the model structure, that places sensors with respect to the class of residual generators that take causal information into account. The algorithms are evaluated on a complex highly nonlinear model of a fuel cell stack system. A number of residual generators that are, by construction, easy to implement are computed and provide full diagnosability performance predicted by the model. [less ▲]

A kurtosis-based dynamic approach to Gaussian mixture modeling

in IEEE Transactions on Systems, Man and Cybernetics. Part A, Systems and Humans (1999), 29(4), 393-399

We address the problem of probability density function estimation using a Gaussian mixture model updated with the expectation-maximization (EM) algorithm. To deal with the case of an unknown number of ... [more ▼]

We address the problem of probability density function estimation using a Gaussian mixture model updated with the expectation-maximization (EM) algorithm. To deal with the case of an unknown number of mixing kernels, we define a new measure for Gaussian mixtures, called total kurtosis, which is based on the weighted sample kurtoses of the kernels. This measure provides an indication of how well the Gaussian mixture fits the data. Then we propose a new dynamic algorithm for Gaussian mixture density estimation which monitors the total kurtosis at each step of the Ehl algorithm in order to decide dynamically on the correct number of kernels and possibly escape from local maxima. We show the potential of our technique in approximating unknown densities through a series of examples with several density estimation problems. [less ▲]