Automated FEM Compliant Tolerance Analysis of Mechanical Assemblies Using Free Body Static Analogy, Machine Learning, and Image Processing

Reaching a high level of precision in mechanical assembly sets requires using a technique that adds up each part’s individual errors in the assembly as a whole. Using force analysis on a free-body diagram, the static analogy approach analyzes assembly problems as a static system to determine the sensitivity coefficients of the assembly. This study presents a new approach to identify, resolve, and derive tolerances from image inputs and their corresponding dimensions. We utilized image processing methods, unsupervised machine learning techniques, and finite element analysis to automate the process. The approach starts by using edge detection algorithms to identify edges and shapes in the input image, followed by a Gaussian filter to remove unnecessary shapes. Coordinates are assigned to important structural points detected in the previous step and processed using Kmeans clustering. The refined data is used to model a structural force analysis, representing the assembly problem as a truss system, which is solved using finite element analysis (FEA) tools to find the sensitivity coefficient. The effectiveness of our approach is validated through a case study, and the method offers a solution for tolerance analysis while preserving comparable precision to the Monte Carlo methodology.

کلیدواژه ها:

Tolerance Analysis ، Finite Element Analysis ، Automation ، Image Processing ، Machine Learning ، Static Analogy