ANALYSIS OF THE STATISTICAL DATA GENERATED BY AN ADAPTIVE STRETCH FORMING PROCESS

: Constant industrial processes improvements represent a fundamental step in the evolution of efficient processing. Due to physical or financial limits, there is a limit to how the mechanical or electronic side can be optimised. A solution for improving industrial processes can come in the form of complex machine algorithms that analyse the process in real-time and decide, with each step, what is optimal. To put this statement into practise, we have designed and implemented a fully operational self-adaptive stretch forming process controlled with the help of a dedicated statistical analysis algorithm. The foundation of the algorithm lies in the deformation theory of metals. In its simplest form, it can be summarised that if a sheet of metal is stretched, its length will increase as the force acting upon it increases until the ultimate tensile strength limit is reached; after this limit, failure occurs. Therefore, the algorithm analyses the material strain controlling the bi-axial nature of the stretch forming process by constantly adjusting for axial force and die speed. It does this through complex computer-vision image analysis techniques for strain measurement and stretching pressure readings as input data. The readings are analysed using the ANOVA method, providing R-squared and p-values for stretching pressure and die speed. The decisions that the algorithm takes are based on the statistical analysis of its previous decision, aiming to improve the overall process R-squared. The overall results are validated by measuring the obtained stretched parts’ deviation to the die shape. Therefore, the measurements were taken using a GOM 3D measuring system. This paper aims to explain the methodology of the algorithm using how the measurements are taken, how the statical analysis generated decisions for controlling the industrial equipment, and to analyse the statistical data generated by the self-adaptive stretch forming algorithm for the experimental study by comparing the decision it takes for each for the 20 processed 1050 aluminium alloys blanks. The results indicate the ideal succession of decisions and which path should be taken to improve the decision-making for both elastic and plastic domains.