Material Characterization Augmented with Artificial Intelligence.

Abstract

Abstract The present paper investigates the application of artificial intelligence to improve the results from simple, non-instrumented, tensile tests, performed with a desktop-size MaCh3D smart universal testing machine. Non-instrumented tensile tests, performed on any testing machine, are affected by both deterministic and random factors that introduce errors in the test results. Specific features of the MaChh3D tester minimize random factors effects on test results while introducing a larger effect of deterministic factors. Artificial intelligence is identified as a novel approach to correct errors in non-instrumented tensile test, capable of simulating a direct strain measure onto the test, replacing traditional contact or non-contact instrumentations (like strain-gages, extensometer and optical measures) that introduce complexity into test procedure and require time for setup. The resulting AI model implementation is described and its performance evaluated in comparison with instrumented tests, also comparing different training strategies. The developed AI-extensometer (artificial intelligence virtual extensometer), is capable of a precise mechanical properties evaluation, with errors from 0 to 10% depending on the specific parameter.