Optimal test generation algorithm with minimum sea

2022-07-26
  • Detail

The optimal test generation algorithm with minimum search space

1 predicts that the test generation problem of digital integrated circuits has always been a NP complete problem. With the expansion of the scale and complexity of integrated circuits, traditional test generation methods have encountered great difficulties in dealing with large-scale circuits under test. Looking for new and effective test generation methods has become an important research topic in the field of integrated circuit testing. In view of the successful application of some optimization algorithms in solving NP complete problems, some scholars in the testing field are trying to turn the test generation problem into an optimization problem. In 1988, chakradhar and others preferred to use Hopfield neural network model to transform the test generation problem into an optimization problem. Using the energy function of the neural network, the test vector corresponding to the fault was calculated by calculating the minimum value of the energy function of the modified asphalt polyethylene tire waterproof coiled material gb18967 ⑵ 003 []. This method is different from the traditional method. It does not need either the fault propagation process or the fallback process; In addition, this method is easy to implement parallel computing. Subsequently, Fujiwara improved chakradhar's neural network model and proposed a three value logic circuit neural network model [3], which expanded the value state of neurons from (0,1) to (0,1,1/2)

in the optimized test generation algorithm, genetic algorithm is often used as the optimization tool. This optimization tool has also been applied in some test generation algorithms of foreign scholars. Some domestic scholars are also there; The optimized test generation algorithm has done some useful exploration

in general, the existing optimal test generation algorithm indirectly expands the search space because it considers some variables in the circuit under test, which limits its practical application. However, its potential parallel computing capability makes the optimized test generation algorithm have broad application prospects. In order to solve the above shortcomings of existing optimization algorithms, this paper proposes an optimization test generation algorithm with minimum search space

2 preparatory knowledge

2.1 genetic algorithm

genetic algorithm is a random optimization algorithm first proposed and studied by John Holland of the University of Michigan in 1975. It simulates the calculation model of Darwin's genetic selection and natural elimination, and finds the optimal solution or quasi optimal solution by continuously optimizing the population in the solution process. As an effective optimization tool, genetic algorithm is simple, versatile, robust and suitable for parallel processing

the main components of genetic algorithm include coding, fitness calculation, parent selection, crossover and mutation. Coding is to convert the points in the parameter space into bit strings, commonly used binary coding and floating-point coding. Fitness calculation is to calculate the fitness value of each individual in the population. Selection is to generate a group of new populations from the current generation. Usually, the selection probability proportional to the individual fitness value is used to select the individual as the parent generation. Crossover is the operation of randomly selecting matching pairs according to a certain probability, and then randomly exchanging a part of two individuals in the matching pair to form a new individual. Variation is the random change of the value of a certain individual with a certain probability

2.2 construction method of evaluation function

evaluation function is the only interface between genetic algorithm and specific application problems. It is a quantitative reflection of the advantages and disadvantages of individuals in the population. Its construction directly affects the efficiency of problem solving

in the optimized test generation algorithm based on genetic algorithm, the evaluation function mainly affects the solution space of the problem. In the existing optimization test generation algorithm, it is usually necessary to determine the assignment of the internal signal lines of the normal circuit. Most Hong Kong citizens want to immediately universal suffrage and the assignment of the internal signal lines of the fault circuit, which expands the search space of the algorithm and reduces the efficiency of the algorithm, which is less affected by the constant extension stress (generally ignored)

for the above reasons, this paper constructs an evaluation function according to the different responses of the basic gate circuit in the normal circuit and the fault circuit, so that it has the minimum search space. In the following description, it should be noted that the fault model adopted by the algorithm in this paper is gate level fault model, the fault type is single fixed type fault, and the circuit under test is combined circuit

the idea of constructing the evaluation function is: select all paths from the fault signal to the original output end. If any signal line in the selected path is regarded as a pseudo output, when the test vector of the fault is applied at the original processing input end, the response of the signal line should be different in the normal circuit and the fault circuit. That is, the output response of the basic gate circuit on the fault signal propagation path is different between the normal circuit and the fault circuit. Accordingly, the following evaluation functions can be constructed:

where n represents the number of basic gates on the fault propagation path in the circuit

the value of function GP (VK, FI) is specified as follows: when VK is input, for the PTH gate, if the value is different between no fault and fault F1, the value of function GP (VK, FI) is 1, otherwise, it is 0. Here, C is a constant

from the construction process of the above evaluation function, it can be seen that using the processing response of the basic gate unit on the fault propagation path to construct the evaluation function can make the algorithm have the minimum search space, which is conducive to improving the efficiency of the algorithm

3 optimized test generation algorithm

in simple genetic algorithm, because the crossover probability and mutation probability are constant, the efficiency is not high, and there is the possibility of "premature". In order to improve efficiency and quickly obtain the optimal solution, this paper adaptively adjusts the crossover probability and mutation probability according to the individual fitness value. When the population tends to fall into the local optimal solution, the crossover probability and mutation probability will be increased accordingly. When the population diverges in the solution space, the crossover probability and mutation probability will be reduced. For an individual with a lower fitness value and a higher fitness value within a period of time, select a lower crossover probability and mutation probability to protect the individual from entering the next generation. For an individual with a lower fitness value, select a higher crossover probability and mutation probability to eliminate the individual. This not only maintains the diversity of the population, but also ensures the convergence of the genetic algorithm, and effectively improves the optimization ability of the genetic algorithm

the test generation algorithm in this paper is as follows: 1) initialize

and randomly select n individuals in the solution space as the initial population

2) evaluate

calculate the fitness value fi of each individual in the population

3) the best preserved individual

the individual with the largest preservation fitness value (Fmax)

4) selection

Copyright © 2011 JIN SHI