Modeling and Analysis of a Wavelet Network-Based Optical Sensor for Vibration Monitoring

In this paper, we present a wavelet network (WN)-based optical position sensor for vibration monitoring using a quad-cell photodetector array. The proposed system uses a He-Ne laser source whose Gaussian beam impinges on the photodetector array. The optical power distribution from the photodetectors is acquired and fed to a position detection system, which maps the power distribution to the x-y position of the laser beam center. The position detection system uses a WN for function approximation to yield the correct mapping. The advantage of the WN is its generalization property. The network can be trained using a small size lookup table and, when properly trained, can provide a correct estimate of the laser position even if the power readings were not included in the training set. An accurate theoretical model of the system was derived to simulate the optical power variation as the beam scans the photodetector array. An experimental setup was next developed to validate the theoretical model. The results obtained show the effectiveness of WNs to estimate the laser beam position for the purpose of vibration monitoring.