|
Abstract Compared with the other traditional mechanical testing methods, indentation method has the advantages of simple manufacturing of samples and in-situ testing. Different from the existing acquisition methods of material mechanical parameters dependent on indentation load-depth curve, this study provides an effective method for inversion of metal plastic mechanical parameters based on residual indentation morphology and neural network learning. Spherical indentation tests?of Cu, Mg and Fe have been carried out by the Instron universal material testing machine, the?residual indentation?morphology of Cu, Mg and Fe was scanned?by?the contour morphology system and the obtained?morphology?feature will?be?used?as?the data?basis?for subsequent?studies. The characteristics of extracted data were analyzed and the corresponding data processing of amplification, rounding, binarization, and high-order digit supplementation were performed. Based on the secondary development of Abaqus software, the residual indentation depth data from numerical simulations with different material parameters were automatically extracted to be used for neural network learning. The activation function, method of initializing neural network parameters, neural network parameters update mode, loss function, optimal parameter finding strategy and neural network structure were compared and selected, ensuring that the neural network learning can achieve a good effect. Based on the residual indentation morphology feature data from indentation test and the neural networks after learning, the plastic mechanical parameters of Cu, Mg and Fe were obtained. Also,?the?related?plastic?mechanical?parameters of Cu, Mg and Fe were?acquired?through?the conventional uniaxial tensile test?and?characterization based on the Instron?universal?material?testing?machine. By comparing?the neural network learning results and?the?tensile?test?characterization?results,?the?relative?errors?of related?plastic?mechanical parameters?of?Cu, Mg and Fe?obtained?from?neural network learning?were?identified, and the effectiveness of the proposed method for obtaining metal plastic mechanical parameters based on neural network learning and residual indentation morphology was verified. The provided method in this study can be extended to the mechanical properties characterization and plastic parameters acquisition of other metal/alloy materials.
|
|
Received: 28 April 2024
Published: 11 October 2024
|
|
|
|
2024, Vol. 45 
