A Novel Unobtrusive Vibration Sensing System for Machine Inspection

Abstract

Spectral vibration signatures of a machine offer one of the earliest indication of a potential failure. The journey from potential failure point to full functional failure is often seen to get traversed in weeks. Thus, it becomes imperative that any condition monitoring system is able to immediately detect anomalous vibration signatures, as they begin to build up. The machines with rotating parts often experience anomalies due to defects like looseness in joints, misalignments, imbalance, mechanical wear and tear etc. Such incidents result in vibrations induced on non-rotating body parts. The preferred mode of vibration sensing is through in situ sensors like accelerometers attached on specific locations of the machine; however a significant benefit (with respect to legacy machines) will be derived if an accurate measurement can be done from a distance, without attaching the sensor on the machine parts. In this regard, we have designed an affordable, unobtrusive and autonomous vibration sensing system, called “ShakeMeter”, where we combine two cost-effective measurement principles namely optical stroboscope with a standard low-frame rate camera and a Doppler sensor. The optical stroboscope detects the vibration frequency with a high degree of precision by capturing the modulo (of division) between the difference between the object’s vibration frequency and the optical frequency. The vibration signal may consist of multiple independent signals. For incorporating such effects, multiple mutually co-prime strobing frequencies are generated and utilized to efficiently estimate the multiple frequency components via Chinese Remainder Theorem (CRT). Experimental results show that our proposed technique can effectively estimate multiple frequencies with a frequency detection error of 0.5% for vibrations occurring up to 1 kHz. While functionalizing this system is an important milestone, its deployment feasibility in real world scenarios is subject to various factors like; sensing distance from the object, luminance conditions, duty cycle of the optical array etc. We study these operational challenges and analyze the performance of this system under those constraints. Based on these empirical studies, we endorse that the system should be operated at a distance of 1–2 m from the vibration scene, and at a strobing duty-cycle rate that does not exceed 1%. Finally, to estimate the vibration signatures induced on a large surface (at multiple points), a Doppler radar is fused with the mentioned technique. Experimental result illustrates that this method can be very beneficial for quickly capturing the spatial vibration information.