Weighing method for truck scale based on neural network with weight-smoothing constraint

Abstract

This paper proposes a novel approach of optimizing a neural network with weight-smoothing constraint (WSCNN) and a weighing method for a truck scale based on WSCNN. In this method, the truck scales’ prior knowledge, i.e., the correlation of the load cells’ outputs, is used to construct the constraint conditions for optimizing a neural network (NN) in the case of a lack of samples, and then the NN’s performance index is constructed and the detail algorithm of WSCNN is given. The experimental results show that the weighing errors of the truck scale with WSCNN are far less than those of NN with data induction method (DIMNN, it is an algorithm of training an NN only by using the data samples, not the prior knowledge). In addition, the WSCNN’s generalization ability is better than that of DIMNN especially in the case of a lack of samples.