Application of Eddy Current Techniques in Electromagnetic Non-Destructive Testing of Metallic Spherical Geometry

Abstract

Metal balls, renowned for their unique mechanical properties such as hardness, durability, and high resistance to wear and corrosion, play a pivotal role in various industries. These spheres find applications in machinery as bearings, valves, and connectors and consumer products like jewelry and game pieces. Ensuring their quality is paramount, given their significance in maintaining the efficiency, safety, and reliability of machinery and products. Inadequate quality can result in mechanical failure and potential accidents. For quality assurance, metal balls undergo tests assessing their radius, electrical conductivity, and magnetic permeability, spotlighting manufacturing defects. However, the lift-off effect introduces errors in thickness measurement using the eddy-current method due to varying distances between the sensor and the test piece. This thesis introduces two strategies to counteract the lift-off effect. Firstly, a method leveraging a linear eddy-current characteristic identifies the ball's diameter without contact. A key insight is the peak frequency feature relating to the lift-off spacing between the coil's center and the ball. Through experimental and analytical results, the radius of metal balls can be discerned with a marginal error below 2%. Secondly, a triple-coil sensor, comprising two receiving and one excitation coil, is proposed for thickness measurement. With this sensor, the difference in peak frequencies of the receiving coils proportionally represents the ball's thickness, minimizing lift-off's impact. By juxtaposing simulation results with experimental findings, a lookup table for different thicknesses is established, with the measurement error confined to 6.1%. Another significant contribution is the model's simplification. In measuring non-magnetic spherical shells' thickness and radius, the original analytical model has been streamlined based on modified spherical Bessel functions' asymptotic forms. This refined model, incorporating only elementary functions, hastens property estimation using the Newton-Raphson optimization method. Both in simulation and experiments, the new model offers accurate estimations in reduced computational time. Lastly, this research extends metal ball quality control techniques to the industrial sector, especially in assessing bearing ball properties which influence rolling bearing's accuracy and longevity. The multi-frequency eddy-current testing technique is employed to estimate bearing ball properties, optimizing the discrepancy between measured and calculated coil inductance spectra through the modified Newton-Raphson algorithm. Moreover, the method's effectiveness is accentuated by simulations and experiments on diverse bearing balls, underscoring its efficacy in industrial application.

Date of Award	31 Dec 2023
Original language	English
Awarding Institution	The University of Manchester
Supervisor	Wuqiang Yang (Supervisor) & Wuliang Yin (Supervisor)