A statistics based Digital Twin for the combined consideration of heat treatment and machining for predicting distortion

This paper introduces a novel concept of Digital Twinning of heat treatment and machining for predicting distortion. A set of physical experiments were conducted, and statistical models based on these trials were created. The experiments involved heat-treating AA7075 billets with multiple input conditions and measuring distortion during machining trials. This trained a Gaussian Process machining model to reproduce the real-life behaviour of a part, and to predict distortions. These predictions matched the shape and magnitude of data points of the trials. The paper suggests further refinements of the model. The developed statistical tool enables distortion prediction to produce right-first-time parts.