Deep Anomaly Detection for CNC Machine Cutting Tool Using Spindle Current Signals

Abstract

In recent years, industrial production has become more and more automated. Machine cutting tool as an important part of industrial production have a large impact on the production efficiency and costs of products. In a real manufacturing process, tool breakage often occurs in an instant without warning, which results a extremely unbalanced ratio of the tool breakage samples to the normal ones. In this case, the traditional supervised learning model can not fit the sample of tool breakage well, which results to inaccurate prediction of tool breakage. In this paper, we use the high precision Hall sensor to collect spindle current data of computer numerical control (CNC). Combining the anomaly detection and deep learning methods, we propose a simple and novel method called CNN-AD to solve the class-imbalance problem in tool breakage prediction. Compared with other prediction algorithms, the proposed method can converge faster and has better accuracy.

FPGA-Embedded Anomaly Detection System for Milling Process

“…In this paper milling process performed with a sharp tool is treated as a negative class (N), and the operation performed with a blunt tool is treated as a positive class (P). Our result is comparable to the state-of-the-art deep convolutional neural network proposed in [10] and generative adversarial networks [9]. The classification results show that all cases of the incorrect milling process have been recognized correctly.…”

“…ANNs, both shallow and deep, have been used for anomaly detection for several years. For example, an effective convolutional neural network-based anomaly detection method to detect tool breakage was proposed in [10]. The procedure was implemented as software, and it was shown that tool breakage of computer numerical control (CNC) machine could be detected by spindle current.…”

A TOPSIS-Assisted Feature Selection Scheme and SOM-Based Anomaly Detection for Milling Tools Under Different Operating Conditions

“…Regarding the anomaly detection-based TCM, most of the researches consider only one set of operating conditions, such as [10] in which DCNN is employed. On the other hand, the authors in [4] applied a DL anomaly detection model based on GAN to detect anomalous tool states under 9 different operating conditions.…”

“…However, considering the fact that performing machining using a worn tool will generate poor quality products and that it is time consuming to acquire labeled data on the shop floor, using a worn tool merely for the sake of collecting measurements could lead to a considerable waste of resources and time. Therefore, relying only on the normal data can represent an efficient and easily applicable implementation of the predictive maintenance [4], [10].…”

Phase Loss Detection Using Current Signals: A Review

“…The hall-effect current sensors are designed as a separate chips or components, which interrupt current lines or use hole-through technologies. They have internal amplification and compensation circuits and output more stable and precise signal [57], which makes implementation of additional protection algorithms possible [58,59], however these sensors need additional space and significantly increase cost of the power converter, which restricts their usage in low-cost drives. Simultaneously, shunt-based current sensors are small, cheap and suitable for low-cost solutions, but their amplification circuits are placed distantly, which increase noise to signal ratio.…”