Statistical analysis of a model based product property control for sheet bending

Fluctuations in semi-finished parts as well as the environmental conditions of a forming process cause uncertainties in the properties of the produced product. Controlling these uncertainties can be realized by using a quality feedback control. Since part properties are usually difficult to measure during a forming process, models of the machine and process characteristics are required. Models derived either from technological considerations or from statistical investigations can be used to predict product properties resulting from a process with specific properties of the semi-finished part. This paper describes the statistical analysis of an adaptive feed-forward model to compensate spring-back in a flexible sheet bending process. It is investigated whether the use of a process integrated measurement of semi-finished part properties, which are fed into the control model, leads to a significant improvement of product quality. The experiments were conducted on a flexible multi-purpose forming machine with the availability of a three-degrees-of-freedom ram motion, including an appropriate tooling system for flexible bending. The process was repeated for each class of the investigated semi-finished parts using the corresponding model parameters in order to generate the data sets for the statistical analysis. A probability density function estimated from the measurement data allows presuming a significant improvement of product quality. By the application of techniques of statistical testing theory, the nonrandomness of the observed improvement is confirmed for the majority of testing fields. This leads to the conclusion, that the quality of forming processes, especially bending processes, can be enhanced by a process control based on a linear regression model considering the properties of the semi-finished part