Research on Multi-bus Automatic Test Technology for Circuit Board Fault Diagnosis

With the development of modern science and technology, the supporting role of testing technology in the development of electronic equipment has become more and more prominent. The construction of test and guarantee equipment has received unprecedented attention. This article describes a multi-bus automation test system consisting of GPIB, VXI, IEEE1394 three buses for functional testing and fault diagnosis of the rice pole circuit. The research on multi-bus automatic test technology for circuit board fault diagnosis was introduced.

The system has the advantages of commonality, easy operation, etc., reduces the test cost, and makes the automatic test of the circuit easier while improving the efficiency of fault diagnosis. With the development of modern science and technology, the supporting role of testing technology in the development of electronic equipment has become more and more prominent. The construction of test and guarantee equipment has received unprecedented attention. In order to ensure that the electronic devices with higher and higher integration in actual applications can work effectively and reliably to obtain more accurate data, a highly efficient test device is required to satisfy the performance test and fault diagnosis of different electronic devices. Therefore, a multi-bus automation test system composed of GPIB, VXI, and IEEE1394 three-buses for function testing and fault diagnosis of rice pole circuits was constructed. The system has the advantages of commonality, easy operation and so on, which reduces the testing cost and makes the automatic testing of the circuit easier while improving the efficiency of fault diagnosis. And using a genetic algorithm-based support vector machine as a classifier to diagnose the data samples obtained from the system test, improving the accuracy of fault diagnosis.

1, multi-bus automation test system

The multi-bus automation test system mainly consists of a test chassis, a main control computer and a digital-analog hybrid interface adapter. The test chassis includes a programmable power supply and a VXI bus chassis.

Test the core part of the chassis - the VXI bus chassis. VXI bus technology has been standardized, modularized and systemized after years of development. It has the advantages of open standards, strong ability of data jogging, high reliability, accurate timing and synchronization, and module reusability. VXI bus chassis consists of dual channel 50MHz arbitrary waveform generator, 6.5 digital multimeter, 2-channel oscilloscope, 64-channel sequential digital I/O, 4-channel A/D, 4-channel D/A, 32-channel relay switch, 8×32 Matrix switches and other modular instruments. VXI modular instruments are small and space-saving

Easy to transport. These performance advantages of VXI are a powerful guarantee for the high accuracy and high operating speed of the test and diagnostic process. The main control computer communicates with the zero slot controller in the VXI bus chassis through the 1394 bus with the advantage of high speed transmission, and is used for global monitoring of the entire automatic test system.

The tester can effectively control the test chassis through the main control computer and complete the test. The main components of the main control computer include a human interface user interface, a digital signal generation module, a test instrument module, a data processing module, a fault dictionary database, a function detection process, a fault diagnosis process, and other control modules. The organic combination of various functional modules can help testers to effectively complete the test work.

The main control computer is connected to the program-controlled power supply via the GPIB bus. The program-controlled power supply includes dual straight flow control source and one-way flow control source. The application of GPIB bus technology is relatively mature, and a large number of test instruments have GPIB interfaces. Through the GPIB interface, several basic instruments and computer instruments can be built into a modular test system, and complex measurements can be performed under the control of a computer. The test system adopts a parallel connection mode, so that the computer can control 3 route control power at the same time to meet various voltage requirements in circuit testing.

The digital-analog hybrid interface adapter is connected to the test chassis, and the test chassis provides the adapter with corresponding power signal and test excitation signal. The adapter communicates with the tested board through the adapter board and feeds back the test data to the test chassis. The test chassis uploads the data to the host computer. In the process of testing, digital probe pens and analog probe pens equipped with digital-analog hybrid interface adapters can be used to test the important intermediate points of the tested board. This structure design facilitates the execution of the test.

The adapter board added between the adapter and the board under test can effectively protect the board under test, preventing the momentary voltage or current caused by human operation errors or test program errors and test instrument failure during the test process. Test board damage or hazard to testers. At the same time, the adapter board can perform corresponding auxiliary operations such as voltage regulation and filtering on the power signal and test excitation signal required for the test of the tested board, thereby improving the quality of the test signal and the reliability of the test data, thereby improving the correctness of fault diagnosis. Sex.

2. Support vector machine classifier based on genetic algorithm

Support vector machine (SVM) first transforms the input space into a high-dimensional space through a nonlinear transformation defined by an inner product function, and then finds the optimal classification surface in this space. Its main idea is to establish a hyperplane as the decision surface. This decision surface can not only correctly classify all training samples, but also make the distance from the closest point to the classification surface in the training sample to be the largest.

Genetic algorithm (GA) is a kind of global optimization algorithm formed by referring to the natural selection and natural genetic mechanism of the biological community. The initial population is generated first, and a group of individuals more suitable for the environment are generated through the operations of selection, crossover and mutation. Through evolution from generation to generation, the population finally converges to a group of individuals most suitable for the environment, and an optimal solution to the problem is obtained.

The basic method of applying genetic algorithms to SVM is as follows:

1) Enter the sample data set, allocate training samples and diagnostic samples, and normalize the sample data.

2) Parameter initialization. That is, some basic parameters are initialized, including the maximum evolutionary generation in the GA algorithm, the maximum number of populations, the probability of crossover and mutation, and the variation range of the parameter C in the penalty function C and the kernel function and the number of cross validations in the SVM.

3) For the parameter C to be optimized and binary coding, and according to its distribution evenly extract some individuals to form an initial population. The average population obtained by uniform extraction is larger than the available information extracted at random and is more advantageous to the algorithm.

4) Set the SVM's fault classification accuracy rate to individual fitness, and the higher the classification accuracy, the greater the individual fitness. According to the size of the individual's fitness value, individuals with larger fitness values ​​are selected from the population to enter the next generation.

5) Perform crossover and mutation operations to form a new generation of populations.

6) When the average fitness value changes continuously less than a certain constant and exceeds a certain number of algebras, the individual with the greatest adaptation is obtained as the optimal solution output, and the optimal solution decoding line obtained is optimized. Otherwise repeat steps 3 to 5.

7) Take the optimization parameters obtained in the previous step as the adopted values ​​of the main parameters C and SVM classifiers for sample training and fault classification.

The SVM with the genetic algorithm not only retains the advantages of the SVM algorithm, but also incorporates the advantages of the genetic algorithm. Starting from the group consisting of multiple points with even distribution, only the fitness value converted from the value of the objective function is needed in the process of seeking the optimal solution, and no other auxiliary information is needed, making the algorithm simpler and not easy to fall into the local optimum. Solution in the dilemma. And avoiding the disadvantages of the difficulty of parameter C selection in the original SVM algorithm, and improving the classification accuracy of the classifier.

3, circuit diagnosis examples

The entire troubleshooting test process can be represented by a flowchart. After the adapter board is correctly installed, the system is powered on. After determining that the test voltage provided by the system is correct, the board to be tested is properly installed and the function test phase is entered. Step by step testing of each functional module, if all the functions passed the test, no abnormal values ​​were detected, then the system shows the circuit board function is normal, the system power off, the test is over. If one or more error values ​​are found during the test, the test system will enter the fault diagnosis test section.

Take the relay switch control circuit with many common components as an example. When a component in the circuit fails, such as an open circuit or an op amp, the voltage amplitude, high value, low value, frequency, and duty cycle of the output signal will change, and the test system will test These circuit characteristic values ​​are saved, used as sample data, and classifiers are used for fault diagnosis.

The test will yield 250 samples, the first 100 will be training samples, and the remaining 150 will be test samples. After the training set and the test set were normalized by data preprocessing, the general support vector machine and the support vector machine based on genetic algorithm were used for fault diagnosis. The diagnostic results are shown in Table 1. The accuracy of the classifier fault diagnosis based on genetic algorithm can reach as high as 99.33%.

4 Conclusion

The multi-bus automation test system integrates a variety of test instruments to facilitate testing. The use of high transmission rate VXI and 1394 bus shortens the test data transmission and processing time during the circuit board test, thereby improving the test efficiency. The system can test different circuit boards through different test and diagnosis procedures, which is versatile and practical, avoids the waste of funds caused by repeated construction of test systems, and reduces the cost of test system development and maintenance. And the support vector machine classifier based on genetic algorithm is added to the test system resource platform, which improves the accuracy of circuit diagnosis.