Weighing Error Compensation for Truck Scales Based on IGWO–LMBP Neural Network

A weighing error compensation method using IGWO–LMBPNN, which combines an improved gray wolf optimization algorithm (IGWO) with the Levenberg–Marquardt backpropagation neural network (LMBPNN), is proposed to enhance truck scale accuracy. It uses output signals from multiple load cells as inputs to the neural network for establishing the error compensation model. The IGWO algorithm is used to initialize the weights and biases of the model, while the LMBPNN algorithm further optimizes these parameters. To evaluate the effectiveness of the IGWO–LMBPNN method, it was compared with the existing compensation methods, including BPNN, LMBPNN, gray wolf optimization with LMBPNN (GWO–LMBPNN), particle swarm optimization with LMBPNN (PSO–LMBPNN), and Fine Gaussian SVM. The experimental results demonstrate a significant improvement in error reduction for truck scale weighing. The mean absolute error (MAE) of IGWO–LMBPNN is 8.5 kg, which was reduced by 38.8%, 29.2%, 19%, 21.3%, and 33.1% compared with BPNN, LMBPNN, GWO–LMBPNN, PSO–LMBPNN, and Fine Gaussian SVM, respectively. These findings not only verify the efficacy of the IGWO–LMBPNN in enhancing weighing accuracy but also highlight its potential for other practical applications.