Application Of Genetic Algorithm In The Particle Inversion

Particle size is the linear scale occupied in space scope of particles. Particle size distribution refers to the percentage of different level of particle size to materials, the narrower particle size distribution, the smaller of the granules dispersion degree is, that is the concentration is higher. The characteristics of powder materials are decided by particle average size, particle size distribution, shape of particles and other important parameters. Therefore it is very important to know particle shape and the distribution of particle average size. Meanwhile the measuring technology is a problem to obtain the characterization of powder materials, and how to compute the particle size distribution fast become is necessary.Light scattering method is one of the widely used ways in current particle measuring applications, and particle inversion is to calculate the particle size distribution measured against scattering intensity from the particles. From the light scattering theory, the relation between light intensity and the particle size on the photoelectric detector can be written the equation: E = TW, to solve W , we must know E and T , this problem belongs to the first kind of Fredholm integral equation in theory, and it is difficult to use of the traditional numerical method to find the answer. With the rapid development of science and technology, particle inversion optimization algorithm constantly updates, and the researchers have been actively looking for all kinds of optimization algorithm to solve the equations.The main work of this paper includes the following aspects:(1) Researching the optimization of application in particle inversion algorithm, including nonnegative least squares and Chahine algorithm.(2) Researching the genetic algorithm in the application of particle inversion. This paper designs a new fitness function, introduces the selection method and reason of genetic operator including coding method and selection operator, crossover operator and mutation operator. By using the genetic algorithm in simulation, in order to reduce the calculation error, we do not choose the binary coding mode which is used widely, the new fitness function of the original algorithm proposed remedy deficiency of the customary error computation. The new fitness function includs global RMS error the maximum error to avoid some bigger error, which makes the inversion result more ideal.(3) Based the nonnegative least squares and Chahine algorithm, the paper made improvement and perfection, which combine genetic algorithm and original inversion algorithm.(4) Based on VB, the paper builds the inversion algorithm platform, which includes each algorithm procedures and synthesizes each optimization algorithm. The paper did the numerical simulation for the single-peak distribution, bimoda distribution and trimodal distribution and mad numerical statistics comparison.The results of computer numerical simulation prove the reliability of the genetic algorithm.The particle size distribution and light spectrum fitting well, and figures of the results obviously reflect the improved genetic algorithm are more. The calculated error of commonly used index is mainly under 1%, and the cumulative curves are also close to the ideal value, which also demonstrates the reliability of the algorithm proposed by this thesis.