Thermal error modeling with dirty and small training sample for the motorized spindle of a precision boring machine

2021

Machines

Thermal error is one of the main sources of machining error of machine tools. Being a key component of the machine tool, the spindle will generate a lot of heat in the machining process and thereby result in a thermal error of itself. Real-time measurement of thermal error will interrupt the machining process. Therefore, this paper presents a machine learning model to estimate the thermal error of the spindle from its feature temperature points. The authors adopt random forests and Gaussian process regression to model the thermal error of the spindle and Pearson correlation coefficients to select the feature temperature points. The result shows that random forests collocating with Pearson correlation coefficients is an efficient and accurate method for the thermal error modeling of the spindle. Its accuracy reaches to 90.49% based on only four feature temperature points—two points at the bearings and two points at the inner housing—and the spindle speed. If the accuracy requirement is not very onerous, one can select just the temperature points of the bearings, because the installation of temperature sensors at these positions is acceptable for the spindle or machine tool manufacture, while the other positions may interfere with the cooling pipeline of the spindle.

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

The Thermal Error Estimation of the Machine Tool Spindle Based on Machine Learning

Chiu

2021

Machines

“…The other type of methods to solve the above problems are the data-driven prediction methods. Two typical and commonly used error modeling prediction methods in this type of method are the neural network prediction method [3][4][5] and the support vector machine for regression (SVR) prediction method [6]. However, the error modeling based on wavelet neural network [3] and LSTM [4] is time-consuming and the BP [5] method is not robust.…”

Section: Introductionmentioning

confidence: 99%

“…However, the error modeling based on wavelet neural network [3] and LSTM [4] is time-consuming and the BP [5] method is not robust. The SVR in [6] just predict time series without establishing specific models for different feed states.…”

Section: Introductionmentioning

confidence: 99%

Multi-Stage Prediction of Feed System Time Series

Shen

Proceedings of CECNet 2021

Liu

2021

In order to reduce the tracking error of the computer numerical control (CNC) feed system and improve the CNC machining accuracy, a novel prediction model is proposed based on fuzzy C-means robust variational echo state network. Firstly, the feed speed time series is clustered, and then reconstructed for different categories. The multi-stage robust prediction models are established to realize the multi-state robust prediction of the CNC machining feed velocity to reduce the tracking error of the feed system. Finally, the reference and actual time series with different feed speed are used to verify the established models. The results show that the proposed method can reduce the tracking error and realize the effective prediction of the time series of the feed system.

“…The analysis shows that in order to avoid over-fitting, certain hyper parameters are defined empirically based on preliminary trials. Although SVR is widely used, it depends on the regulation of super parameters [15]. The kernel function must satisfy Mercer criterion.…”

Section: Introductionmentioning

confidence: 99%

A Hybrid Prognostics Approach for Motorized Spindle-Tool Holder Remaining Useful Life Prediction

Han

Advances in Asset Management and Condition Monitoring

Qiu

et al. 2020

The quality and efficiency of high-speed machining are restricted by the matching performance of the motorized spindle-tool holder. In high speed cutting process, the mating surface is subjected to alternating torque, repeated clamping wear and centrifugal force, which results in serious degradation of mating performance. Therefore, for the purpose of the optimum maintenance time, periodic evaluation and prediction of remaining useful life (RUL) should be carried out. Firstly, the mapping model between the current of the motorized spindle and matching performance was extracted, and the degradation characteristics of spindle-tool holder were emphatically analyzed. After the original current is de-noised by an adaptive threshold function, the extent of degradation was identified by the amplitudes of wavelet packet entropy. A hybrid prognostics combining Relevance Vector Machine (RVM) i.e AI-model with power regression i,e. statistical model was proposed to predict the RUL. Finally, the proposed scheme was verified based on a motorized spindle reliability test platform. The experimental results show that the current signal processing method based on wavelet packet and entropy can reflect the change of the degradation characteristics sensitively. Compared with other two similar models, the hybrid model proposed can accurately predict the RUL. This model is suitable for complex and high reliability equipment when Condition Monitoring (CM) data is scarcer.