Basic Principles of the Expert System for Solder Joint Quality Analysis and Evaluation
There is an inevitable relationship between SMT solder joint shape and solder joint quality, solder joint structure and process parameters. The quality of solder joints can be compared, analyzed and evaluated through the design expert system. In the analysis and evaluation expert system, expert knowledge is generally obtained through the experience of experts in the SMT field. Therefore, there are large limitations and uncertainties in the acquisition of knowledge. Moreover, the solder joint quality analysis and evaluation expert system itself is a rule library with many influencing factors. It is a very large system with complex reasoning process, so it is difficult for the traditional expert system to achieve the speed and accuracy of the quality analysis and evaluation of all types of solder joints. For this reason, technologies and methods such as expert systems based on fuzzy control and expert systems based on neural networks can be adopted for different analysis and evaluation objects.
The schematic diagram of the structure model of the SMT solder joint quality analysis and evaluation expert system based on neural network. The basic principle of the system is: use the knowledge of experts in the SMT domain to train and test the neural network in advance, and store the obtained reasonable connection weights in the database in the form of knowledge as the basis for the expert system’s reasoning. When working, the result of comparing the actual solder joint shape parameter with the reasonable solder joint shape parameter (ie the deviation between the two) is used as the system input, and the expert system is used for reasoning. The reasoning result (the output of the system) is a set of assembly process parameters Adjust the value, the control interface module of the system or the process engineer can adjust the assembly process parameters according to the result, or take other corresponding troubleshooting measures.
