Abstract: Most of non-linear type one and type two control systems suffers from lack of detectability when model based techniques are applied on FDI tasks. This research is centered on a strategy based on Closed Loop Frequency Response Test (CLFRT) to estimate plant parameters which includes massive neural networks based functional approximation procedures. Nominal plant parameters are matched against on-line estimated parameters on a parity space approach. The strategy to carry out this task consists in Developing a fault tolerant data acquisition strategy to achieve a database to be used in neural networks training. Proposing and implementing a methodology to estimate plant parameters by functional approximation based on backpropagation neural networks.