This work presents an algebraic approach to fault diagnosis for robots, utilizing polynomial approximation to detect and identify faults in real time based solely on measurement data. Two distinct fault types are considered: unknown signal faults and state transition faults, each addressed with dedicated procedures. The method takes external disturbances, parameter uncertainties, quantization effects, and numerical errors into account. Analytical expressions for error bounds and systematic rules for the selection of design parameters are derived to support practical implementation. The approach does not rely on full state reconstruction or detailed observer structures, but instead evaluates differential-algebraic expressions using the polynomial approximation operator. The proposed framework is validated through extensive simulations and experimental studies on industrial robot platforms, including the application to industrial robots. Furthermore, an extension to general nonlinear systems is presented, demonstrating the broader applicability of the method beyond rigid-body dynamics. The results contribute to the development of robust, model-based fault diagnosis methods suitable for integration into robotic systems with functional safety requirements.
Alexander Lomakin (Wed,) studied this question.