Prediction of Transient Thermal Deformation in Press by Detrending Time-Series Variations

The servo presses are convinced that they can provide more flexibility in product design with a more efficient process. In the context of a future where precision machining is becoming increasingly critical, machining tolerances caused by thermal deformation and the potential risk of damage to molds must be considered seriously. This study proposes a feasible method for estimating the deviations of the clamping positions by measuring temperatures at key points. We first use finite element analysis to obtain the simulated thermal deformation behaviors of components in the press. Combined with our innovative detrending time-dependent approach, the Energy Influence Depth (EID) proposed in this study is used to compute the depth to which energy can be transmitted within a predefined area. Then, we train a compact prediction model to calculate ideal-state clamping position deviations under the thermal effect, achieving an accuracy of over 96% and 97% for the two sides of the moving head, which indirectly validates the effectiveness of our proposed Energy Influence Depth. Next, to obtain the experiment-based prediction model, we then conduct thermal-deformation experiments on the actual machine. Through the same model development method as the ideal-state model, we use the experimental data to obtain the EID value and train an experimental position deviation prediction model, with the test normalized root-mean-square error of 0.1 and 0.07. In conclusion, the model predicts the trend of the experimental results consistently. Although in our tests, sometimes, the model overestimates or underestimates the bed position deviation, which might be due to the lack of training data. Nevertheless, the method proposed in this study, combining the energy influence depth with a compact model to obtain the bed position deviation prediction model, is both feasible and practical.