Thermal error modeling based on BiLSTM deep learning for CNC machine tool

Abstract

The machining accuracy of computer numerical control machine tools has always been a focus of the manufacturing industry. Among all errors, thermal error affects the machining accuracy considerably. Because of the significant impact of Industry 4.0 on machine tools, existing thermal error modeling methods have encountered unprecedented challenges in terms of model complexity and capability of dealing with a large number of time series data. A thermal error modeling method is proposed based on bidirectional long short-term memory (BiLSTM) deep learning, which has good learning ability and a strong capability to handle a large group of dynamic data. A four-layer model framework that includes BiLSTM, a feedforward neural network, and the max pooling is constructed. An elaborately designed algorithm is proposed for better and faster model training. The window length of the input sequence is selected based on the phase space reconstruction of the time series. The model prediction accuracy and model robustness were verified experimentally by three validation tests in which thermal errors predicted by the proposed model were compensated for real workpiece cutting. The average depth variation of the workpiece was reduced from approximately 50 µm to less than 2 µm after compensation. The reduction in maximum depth variation was more than 85%. The proposed model was proved to be feasible and effective for improving machining accuracy significantly.

Application of Machine Learning in the Precise and Cost-Effective Self-Compensation of the Thermal Errors of CNC Machine Tools – A Review

“…Artificial neural networks (ANNs) are more complex models which are difficult to implement on the CNC controller. Therefore an industrial computer is used to implement them [23]. The ANNs most often used in thermal error modelling are: a multilayer perceptron (MLP) [4,6,7,9,14,15,17,21,24,26,29,33,35,[40][41][42][43][44] and a radial basis function (RBF) network [8,9,16,19,20,27,28,34,36,38].…”

“…It is demonstrated in section 5 that ANNs showed higher accuracy (in μm and %) than linear models. Thanks to thermal error compensation based on a machine learning model the machining error was reduced by at least 30% [10,11,14,15,17,23,26,32].…”

“…It appears that most often a few or 10-20 sensors have been used. In the most accurate thermal error modelling methods the following numbers of sensors were used: 12 temperature sensors [20], 8 temperature sensors [23] and 6 temperature sensors and an infrared camera [10]. In order to determine a sufficient number of sensors, we divided the modelling methods into three groups using: a few, 10-20 and more than 20 temperature sensors.…”

“…The thermal error has also been modelled as a function of: an infrared camera image [6,10], temperature and spindle rotational speed measurements [13,23] or measurements of temperatures and deformations of specific machine parts [7]. The rotational speed of the spindle has a direct effect on the heating of specific machine tool structure components.…”

“…Since the overall precision of thermal error compensation depends on the modelling accuracy and the thermal error measurement precision, it is important that the measurement of the thermal error be very precise. The thermal error has been measured by means of: an inductive sensor [13,16,17,21,23,25,26,30,[33][34][35][36][37][38]42], a capacitive sensor [14], a laser interferometer [9,19,31] and a laser reflective sensor [7,24,43,44]. The thermal error measurement procedures used in 5-axis machine tools have been: the R-test [46] and ETVE [25].…”