Improved BP Arithmetic in Moisture Content Measurement with Microwave Resonant

Traditional linear regression is the primary factor that affects measurement precision in measuring moisture content with microwave resonator. A regression is put forward based on an improved BP algorithm to modify the measurement result. First, the regression neural network is pre optimized by using the macro search ability, parallel operation and strong robustness of genetic algorithm. Then, integrating the gradient descent method of BP algorithm, the presented algorithm can effectively avoid the traditional BP algorithm of falling into local minimum, at the same time, high prediction accuracy and fast convergence speed are maintained. It has the characteristics of global superiority and accuracy for optimization, thus improving the measurement accuracy. The experimental results show that the mean square error between predicted moisture and actual moisture is 0.0109, the average absolute error is 0.0702, the average relative error is 0.1161, and the determination coefficient is 0.9989.