A data fusion method based on an optimized neural network for MEMS accelerometer arrays

In the domain of precision measurement, achieving high precision and extensive dynamic range detection for unique vibration signals is not readily accomplished by relying on a single MEMS accelerometer. In this paper, a data fusion method based on an optimized neural network for MEMS accelerometer arrays is proposed. To solve the problem of spatiotemporal misalignment of the multichannel outputs, the techniques of twelve-position calibration and cubic spline interpolation are employed. Subsequently, the data is processed through fusion training using the BP neural network model. However, traditional neural network suffers from slow convergence and fall into local optima. To address these issues, the initial weights and thresholds of the BP neural network are refined using an Enhanced Whale Optimization Algorithm (EWOA-BPNN), which enhances the network’s convergence properties and global search capability. The experimental results show that the proposed optimized neural network algorithm achieves data fusion of accelerometer arrays, and the equivalent dynamic range extends to 140 dB. This represents a remarkable improvement in performance compared to single-channel MEMS accelerometers.