Scholarlink

<p>Joint clearances in mechanical systems arise inevitably from manufacturing tolerances, assembly requirements, and progressive wear. However, their detection and prediction remain challenging. This paper presents a novel integrated framework that combines kinematic analysis, dynamic simulation, and machine learning. The framework predicts joint clearance from in planar four bar mechanism from coupler curve deviations, which are directly measurable and do not require system disassembly. A rigorous theoretical foundation is established using loop-closure equations and coordinate transformation matrices. This foundation characterizes clearance induced deviations in four bar mechanisms. Extensive computational simulations were conducted across 100–2000 RPM and 0.01–2.5 mm clearance ranges. These simulations generate 22 geometric and statistical features from the deviation curves. Through systematic feature engineering, including speed normalization and linear scaling, parametric regression models are shown to achieve superior generalization performance. Specifically, polynomial regression achieves <InlineEquation ID="IEq1"> <EquationSource Format="TEX">\(R^2 = 0.9986\)</EquationSource> <EquationSource Format="MATHML"><math> <mrow> <msup> <mi>R</mi> <mn>2</mn> </msup> <mo>=</mo> <mn>0.9986</mn> </mrow> </math></EquationSource> </InlineEquation> under extrapolation conditions. Experimental validation on a physical four bar mechanism confirms simulation accuracy. This establishes the feasibility of clearing the prediction from kinematic data alone. The proposed methodology facilitates a basic framework for cost-effective condition-based monitoring and predictive maintenance for planar mechanisms that operate in a similar manner. It does not require direct wear measurement and offers significant practical utility for industrial applications and bearing life management.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Machine learning-based joint clearance prediction from coupler curve deviations in planar four bar mechanism

  • 期刊论文

  • 发表日期：

    2026年5月5日

查看全文
  • Dixitkumar Gohil,
  • Anirban Guha,
  • Makarand Kulkarni

摘要

Joint clearances in mechanical systems arise inevitably from manufacturing tolerances, assembly requirements, and progressive wear. However, their detection and prediction remain challenging. This paper presents a novel integrated framework that combines kinematic analysis, dynamic simulation, and machine learning. The framework predicts joint clearance from in planar four bar mechanism from coupler curve deviations, which are directly measurable and do not require system disassembly. A rigorous theoretical foundation is established using loop-closure equations and coordinate transformation matrices. This foundation characterizes clearance induced deviations in four bar mechanisms. Extensive computational simulations were conducted across 100–2000 RPM and 0.01–2.5 mm clearance ranges. These simulations generate 22 geometric and statistical features from the deviation curves. Through systematic feature engineering, including speed normalization and linear scaling, parametric regression models are shown to achieve superior generalization performance. Specifically, polynomial regression achieves \(R^2 = 0.9986\) R 2 = 0.9986 under extrapolation conditions. Experimental validation on a physical four bar mechanism confirms simulation accuracy. This establishes the feasibility of clearing the prediction from kinematic data alone. The proposed methodology facilitates a basic framework for cost-effective condition-based monitoring and predictive maintenance for planar mechanisms that operate in a similar manner. It does not require direct wear measurement and offers significant practical utility for industrial applications and bearing life management.