Mathematical Balance
Mathematical configurations containing more independent measurement equations than there are unknown parameters allow for the reduction of individual sensor error. Creating an over-determined system enables multi-sensor arrays to calculate the most probable value of a physical variable with high accuracy. This configuration provides redundant data points that can be processed using statistical techniques to identify and discard faulty sensor readings.
Sensor Fusion
Multiple independent transducers measure the same physical phenomenon from different physical locations or through different methods. The output from these sensors is combined in a central processor to resolve the true value. Because the system has more data points than necessary, it can continue to operate accurately even if one of the sensors fails completely.
Diagnostic Verification
Residual analysis of the measurement equations highlights discrepancies between individual sensor channels. When the residual values exceed a predetermined threshold, the system flags a potential sensor failure or drift. This continuous monitoring improves the safety of industrial processes.
Calibration Algorithm
Statistical algorithms process the redundant data to update the calibration coefficients of the individual sensors during operation. The software adjusts the weights assigned to each sensor based on its historical performance. This automatic correction maintains system accuracy over time.
