A process for the detection of deviations of the surface shape of a predetermined shape by means of a three-dimensional measurement acquisition and a subsequent processing with an artificial neuronal network

摘要

With the aid of optical and mechanical measuring method, the works can be checked for their quality. In the simplest case, measured in this case, firstly a masterpiece. The statement about the quality of the part to be measured can be ascertained by comparing (difference) take place with the masterpiece. Dollars a, the measuring a master part and subtraction of the original part is only possible, however, if for all the parts of a correspondingly precise adjustment can be realized. Dollars a are the errors in the comparison to the size of the measuring part, however, very small, increases the demands on the adjustment. Often there is an adjustment with the necessary accuracy is no longer possible. In most cases, the parts not only displaced, but is bent in themselves and also rotated. In this case it is no longer possible, by difference formation of a master unit with the measuring part to detect automatically error. Dollars a of this problem is solved with the invention the following way: dollars a the object to be measured is measured. The measured values are in a modified content addressable memory (artificial neural network) and are called up at the output (recall). In the weights of the modified associative memory are one or more master parts stored in different positions and locations. As a result, the entire bandwidth of the possible variations in the detected master parts. Dollars a at the output of the modified associative memory only those components of the measurement data, which the stored ..