Ubiquitous Domain Adaptation at the Edge for Vibration-Based Machine Status Monitoring

Abstract

This paper presents a Ubiquitous Domain Adaptation (UDA) and generalizability technique for vibrationbased automated machine status monitoring at the edge. The method significantly reduces the effects of signal noise artifacts and device/usage-specific vibration signatures using basic timefrequency domain signal operations and a lightweight ensemble of data-driven classifiers, allowing the method to be used for reliable domain-invariant status monitoring of motorized equipment. An experimental setup using vibration data from an air-cooled electric blender motor (source-domain) is used to train an automated machine state identification classifier that can identify the operating states of an eccentric rotating mass vibration motor (target-domain). Initial deployment of this method on target-domain motorized devices resulted in a machine status monitoring accuracy of at least 81.6% and a maximum training accuracy of almost 99% on known data of the source-domain and 91.49% for unseen data in the target-domain within an acceptable time frame. The performance of the proposed method is also comparable across platforms ranging from resource-constrained edge to a resource-rich cloud. This approach facilitates the use of noisy or uncalibrated sensor data in data-driven machine status monitoring tasks, therefore allowing for the development of reusable, low-cost monitoring systems that require meagre developmental effort, resulting in accelerated deployment times.

Platform and Hardware Independent Reliable Signal Processing

“…Hence, only those frequencies were analysed. In contrast, the entire frequency range of the speech and vibration signal were analysed as consonants (i.e., "s", "h", and "f") are in the high-frequency range [41] and vibrations often occur at high frequencies equal to the machine's rotational speed [20]. The benchmark was performed on an eight-core 2.30-GHz CPU running macOS Big Sur.…”

“…However, it has been designed to function as a versatile platform for CWT-based analysis of various signals. In the future, signal-specific analysis modules could be implemented to provide researchers with tools such as phoneme recognition or noise removal for speech analysis [21], peak detection [42], filtering or heart rate variability extraction [43] for ECG research, diagnostic reporting on vibration data [20], or ridge extraction [44] for general purposes. Furthermore, other CWT-based time-frequency algorithms such as the Synchrosqueezed Transform (SST) [45] or Superlet Transform (SLT) [46] could be implemented to improve accuracy even more.…”

“…Simultaneously, without high precision, safety, reliability, and security are no longer guaranteed. Minor but important changes in heart rhythm [19] or machine vibration [4,20] would go unnoticed, speech analysis fails [21], and security becomes an issue in radio frequency-based authentication of IoT devices [22].…”

An In-Depth Study of Vibration Sensors for Condition Monitoring

“…An accelerometer collected vibrations from both good and defective gears in the gearbox, and tests showed distinct features that validated the model's accuracy. An IMU sensor and current/voltage sensors were used in the vibration monitoring system by Mukherjee et al [94], and data were sent to an Arduino-based ESP-32 board. A Raspberry Pi gateway gathers and analyzes these data using MQTT for reliability.…”

“…Vos et al [133] and Mauricio et al [142] acquired data from helicopter gearboxes to monitor these systems' conditions. Data from a rolling mill gearbox was used by Lu et al [122], data from a blender motor gearbox was collected by Mukherjee et al [94], and data from a wind turbine gearbox was used by Amin et al [41]. Some articles, such as those by Ong et al [45], Peters et al [143], Wang et al [134], Civera et al [91], Inturi et al [93], Mazzoleni et al [95], Samant et al [4], and Krot et al [101] did not mention the gearbox type.…”