Weighing Fusion Method for Truck Scales Based on Prior Knowledge and Neural Network Ensembles

IEEE Trans. Instrum. Meas.

et al. 2015

Self Cite

This paper presents a new approach for compensating nonlinear errors of a load cell, based on the optimal neural network with an augmented Lagrange multiplier (ALMNN). The load cell has serious nonlinear errors, which lower the accuracy of the weighing results. In this proposed approach, first, we construct the constraints for training the neural network (NN) using the load cell's prior knowledge, i.e., the monotonic increasing property of a load cell's input-output function. Second, we create the augmented Lagrange function with a multiplier and a penalty factor to serve as an NN's performance index, and then deduce the ALMNN's detailed training algorithm. The penalty factor plays an important role in ALMNN, thus the reasonable range of numerical value is discussed in this paper. Meanwhile, the weighing principle of a strain gauge load cell is introduced and its nonlinear error model is established. Finally, the experimental results demonstrate the effectiveness of ALMNN. By comparing with the conventional data induction method (DINN, i.e., the training of NN not with the prior knowledge but only the data samples), ALMNN significantly improves the NN's generalization ability, especially in the case of lacking training samples. In addition, the applicability of ALMNN is verified by three additional experiments.Index Terms-First-order derivative constraint, Lagrange multiplier, load cell, monotonicity, neural networks (NNs), nonlinear error compensation.

show abstract

Section: G Discussion Of Almnn's Expansive Applicationsmentioning

confidence: 99%

Nonlinear Error Compensation for Load Cells Based on the Optimal Neural Network With an Augmented Lagrange Multiplier

IEEE Trans. Instrum. Meas.

et al. 2015

Self Cite

show abstract

“…We constructed the weighing model of truck scale by using RBRNN, PKNNEs (Lin et al, 2014a), support vector regression (SVR) and PFNN with 30 and 60 training samples, respectively, and the comparative results are shown in Figure 14. In RBFNN, the number of hidden neurons M is 15 and the BP is applied in training this RBFNN.…”

Section: Comparison Of Pfnn Rbfnn Pknnes and Svrmentioning

confidence: 99%

“…into the standard NN, and proposed a novel generalized-constraints NN (GCNN), which has better approximation ability. In particular, Lin et al (2014a) proposed a weighing fusion method for a truck scale based on prior knowledge and NN ensembles (PKNNEs). In this proposed PKNNEs, the ideal weighing model of a truck scale is regarded as prior knowledge and embedded into the NN's performance index, which can improve the NN's generalization ability.…”

Section: Introductionmentioning

confidence: 99%

Weighing method for a truck scale based on optional neural network with penalty function

Transactions of the Institute of Measurement and Control

et al. 2016

Self Cite

This paper presents a weighing method for a truck scale based on a neural network (NN) with penalty function (PFNN). Firstly, the truck scale’s prior knowledge, i.e. the positive partial derivatives of the truck scale’s input–output function and the distributions of truck scale’s permissible weighing errors, is used to construct the constraints for optimizing the NN. Then, the penalty function is applied to construct the new NN’s performance index, and the detailed algorithm for training this NN is given. Finally, the method for assigning the value of the penalty factors is also discussed. The comparative experimental results show that the PFNN’s generalization ability is better than that of a data-induced NN (DINN) especially with a lack of training samples (the DINN is a method for training an NN only by using the data samples, not prior knowledge), and the weighing errors of the truck scale with PFNN are far less than those of DINN. In addition, the convergence of the PFNN is faster than that of the DINN.

show abstract

“…Fortunately, the prior knowledge is very useful for optimizing NNs [13][14][15][16][17][18][19]. Yajun et al [14] used some obvious prior knowledge, such as the symmetry, the ranking list, the boundary and the monotonicity, to propose a constrained neural network regression model for improving the conventional NN's generalization ability.…”

Section: Introductionmentioning

confidence: 99%

“…Lou et al [18], discussed a popular and effective method that embeds system's prior knowledge into neural networks, the prior knowledge is including invariance, monotonicity, homogeneity, concavity, etc. Lin et al [19] proposed a weighing fusion method for truck scale based on prior knowledge (e.g., the ideal weighing model of a truck scale) and neural network ensembles. These aforementioned methods are useful examples for optimizing the neural network when the training samples are lacking.…”

Section: Introductionmentioning

confidence: 99%

Weighing fusion method for truck scale based on an optimal neural network with derivative constraints and a lagrange multiplier

et al. 2015

Measurement

Self Cite