Sheet metal forming global control system based on artificial vision system and force–acoustic sensors

Forming processes are manufacturing processes based on deformation of raw material applying pressure in one or several stages until getting the final product. This process depends on many factors, e.g. process parameters, material properties or lubrication, leading to possible defective parts. Correct forming of parts is very important as any defective part may result in big economical losses, e.g. the return of a complete set of parts or the loss of some clients. Thus, in our European Craft Pro2Control project, leading German, French, Italian and Spanish companies, universities and forming industries are defining and implementing a zero-defect forming control system, minimizing costs and maximizing the throughput of parts.Commonly integrated sensors (force and acoustic) do not allow to detect all types of faults in this kind of applications. Thus, we consider a multisensor approach, associating artificial vision to the previous sensors. Vision may allow better understanding and/or characterizing of force and acoustic measures, while it can detect faults that the other sensors may not.The design of the artificial vision system has been separated into a pre- and post-processing part. The former consists of a proprietary intelligent camera built up by one of the project leaders (Delta Technologies Sud-Ouest or DTSO), which contains a CMOS sensor, FPGAs, RAMs and a USB 2.0 connection, and where parallelizable bottlenecked image-processing algorithms are implemented (whenever image-processing algorithms can be implemented on FPGAs, processing times will be about 100 times faster than on a standard microprocessor). The latter consists of proprietary non-bottlenecked image-processing algorithms, implemented on a regular PC, using open source libraries. This approach provides a fully mastered development, and guarantees durability and maintainability of the system (non-dependence on commercial items), as well as a new scale of production throughputs. The system is customizable, and the multisensor approach will improve fault detection robustness.