Study of the Fault Diagnosis Model of High Pressure Roller Mill Gearbox Lubrication System

The fault phenomenon of high pressure roller mill gearbox lubrication system is not easy to be find in many cases, the fault of system is easy to be ignored, and it is more difficult to judge with the traditional method. For this reason, the fault diagnosis model of the particle swarm neural network has been established by using actual sample data, determining the cause of fault through the actual monitoring data. The practice has proved that it has better prediction effect. Keywords: BP network, fault diagnosis, high pressure roller mill, particle swarm algorithm. 1. INTRODUCTION The high pressure roller mill is widely used in mining, and cement enterprises etc. As an important equipment of crushing, it directly affects the production and benefit of enterprises. It is well received by enterprise because of its characteristics of energy saving and high efficiency, and has been widely promoted. As the important transmission mechanism, the high pressure roller mill gearbox directly affects the normal operation of the equipment, but its lubrication system is the assurance of safe and reliable production. Whether the lubrication system is working or the problem can be found quickly. Whether the cause of the problem is quickly identified for troubleshooting. These questions have become a barrier in the daily inspection and maintenance. But in practical applications, because of limitations of the technical level of the operator, it can not be quickly found when the fault occurs, which leads to equipment damage, due to which the production can't continue. The situation thus exerts great economic losses and is not conducive to the repair and maintenance of the equipment, and resultantly it can bring a lot of trouble for the production of the enterprise. So, finding a fast and effective diagnosis method to identify ahead the fault of lubrication system, so that it can be treated by the maintenance personnel, has thus become a very valuable subject. For this, the paper presented the method of particle swarm neural network fault diagnosis, which is established based on the actual data. The paper has diagnosed and predicted the fault by using the actual data, and the practice showed that prediction result is very good. 2. PARTICLE SWARM OPTIMIZATION ALGORITHM 2.1. The Basic Principle of Particle Swarm Optimization Algorithm The particle swarm optimization algorithm is the result of the study of birds prey behavior. It is a kind of optimization *Address correspondence to this author at the School of Physical and Electrical Information, Luoyang Normal University, Luoyang 471022, Henan, China; Tel: +8615937912295; E-mail: [email protected] tool based on iteration. Firstly, it is initialized to a group of random particles (random solution) when the optimization began, and then it can find the optimal solution through iteration. In each iteration, the particles can update themselves by tracking between two extremes. The first extreme value is the optimal solution currently found in the whole population. The extreme is the global extremumgBest. In addition, we need not the whole population, but only parts as the neighbor of particles, then the extreme in all the neighbors is the local extremum-1Best. The second extreme value is the optimal solution found by the particle itself, which is known as the individual extremum-pBest. The position and speed of random initialization of particles composed of the initial population is calculated by the algorithm. The initial population is evenly distributed in the solution space. The position and speed of the ith particle in the n-dimensional solution space can be respectively represented as Xi = ( Xi1, Xi2,, Xid ) and Vi = (Vi1,Vi2,,Vid ), to find the optimal solution through iteration. In each iteration, the particle update occurs at own speed and position by tracking the two extremes. One extreme is the optimal solutions so far found about particle itself, which is called the individual extremum P = (P1, P2,, Pid ), that is pBest. Another extreme is the optimal solutions so far found about the particles' neighborhood, which is called the optimal particles Pg = (Pg1, Pg2,, Pgd ) found throughout the neighborhood, that is gBest [1, 2]. According to the formulas (1) and (2), the particles update occurs at own speed and position: