Article Preview
TopIntroduction
The utilization of industrial control and factory automation has gained prominence recently as a strategy for addressing labor shortages and escalating costs. This trend has precipitated an increased demand for advanced factory equipment and methodologies capable of predicting equipment failures and executing maintenance procedures. When equipment malfunctions, organizations dependent on it for service provision or product manufacturing incur financial losses. Additionally, equipment repair can be costly, underscoring the importance of early problem detection and preventive maintenance to mitigate breakdowns and minimize economic impact.
Malfunctions in machines are often identified through manual inspection or the installation of physical contact sensors, which gather data such as voltage and current changes, machine vibrations, and other relevant information to detect issues. However, the use of human labor for this purpose is not suitable for factory automation, and the installation of physical sensors and cameras requires additional cost and time. Therefore, a noncontact sensor system that can measure machine movements without directly attaching sensors to machine tools is necessary. Noncontact sensors, such as microphones, optical fibers, and cameras, can detect abnormal machine states. However, each sensor has limitations in terms of installation positions, requiring a deeper understanding and practical knowledge of the equipment. Wi-Fi sensing is a technology that utilizes changes in Wi-Fi radio wave channel state information (CSI) to function as a noncontact sensor for invisible areas. Unlike cameras, it can detect changes in the surrounding environment; therefore, it is effective for detecting machine malfunctions in areas that are difficult to observe. Wi-Fi sensing technology offers a cost-effective alternative to conventional contact sensors, providing a reassuring solution for predictive maintenance in industrial settings.
Our research aimed to develop a system that utilizes a noncontact sensor with Wi-Fi sensing capabilities to detect machine vibration and predict impending machine failure. However, it is difficult to predict machine tool failures by relying solely on Wi-Fi sensing. To address this limitation, this research extracted information about the malfunction and measurement environment from the description section of daily work reports, which record the operating state of the machine tool. This information is combined with the sensing results to provide insight into the operating state of the machine and to interpret machine tool failures. However, to implement a predictive maintenance system for machine tools, it is essential to estimate the operating state of the machine tool using Wi-Fi sensing in a noncontact manner. Wi-Fi sensing measures changes in the transmission path state and is thus highly susceptible to alterations in the surrounding environment. Moreover, in industrial settings, additional factors, such as the movement of manufactured goods and machine tools, affect Wi-Fi communication. This research hypothesized that the frequency of vibration and changes in the surrounding environment differs significantly and aimed to establish a novel approach that mitigates the effects of the surrounding environment using techniques such as frequency components and signal separation.
The extent to which the operating state of a machine tool can be estimated using Wi-Fi sensing remains unclear. Therefore, this study conducted a preliminary evaluation of Wi-Fi sensing. Due to the challenge of accurately reproducing the failure state of a machine tool, this study utilized a numerical control (NC) machine tool. By programming the operation of an NC machine tool, the operating state of material processing can be modified. This study aimed to determine if it is possible to distinguish between the stopped, idling, and cutting states of an NC machine tool using Wi-Fi sensing based on the different vibrations occurring in these states. In addition, the evaluation included an assessment of whether it would be possible to measure the differences in machining vibration during cutting by varying the object being cut. For these states, the feasibility of estimating the operating state of the NC machine tool and the machining state of the object being machined was evaluated by analyzing the transmission path information acquired from the Wi-Fi module as CSI.
This paper, in the second section, introduces related research and then, in the third section, describes Wi-Fi sensing using CSI. Section 4 outlines the experimental environment utilized in this study, and Section 5 assesses the preliminary evaluation of machine vibration detection through CSI. Finally, Section 6 presents the conclusions of this study.