Nonlinear Errors Correction of Pressure Sensor Based on BP Neural Network

BP neutral network and its improved algorithms are applied to compensate sensor's performance. The defects of BP, for example, converging slowly, being easy to converge to minimum of one part are improved efficiently. Training programs are done. Results show that the performance of sensor is improved highly. Network has a high converging speed and good precision. The correction precision increases with the increasing number of nodes in the hidden layer. When the number of nodes in the hidden layer is 18 and the neural network model converges in average 28 iterations, the Error Index is less than 10
^-3
.