The application of numerical control in the rectification of the valve problem of the rotary air compressor frame

This paper applies the most widely used fuzzy C-means clustering method. The basic principle is as follows: the data sample set X=x1, x2...xn" contains n samples, which are to be classified into C class, and the membership degree of any kth sample in X to the i-th class is uik. The classification result can be used. The fuzzy matrix U (membership function) indicates that U=(uik)∈μc×n, and three conditions must be met: (1) ci=1”uik=1, Ak; (2)uik∈[0,1] (3)0

Jm(U,V)=nk=1"ci=1"(uik)m‖xk-Vi‖2,1≤m≤∞(1) where V=(Vi), i=1,2,..., c, ‖‖ is an arbitrary vector norm of the Rp space. When m=1, uik∈[0,1], it becomes a general C-means clustering method. At the same time, a proper fuzzy C-group classification matrix U and an appropriate cluster center V can be found by weighted least squares method to minimize Jm(U, V). This problem can be attributed to the conditional extreme value problem of function (1) under the constraint ci=1"uik=1. With Lagrange multiplier method, if m>1, xk≠Vk, it can be proved: uik=1cj=1∑ ‖xk-Vi‖‖xk-Vj‖2m-1, Vi=nk=1∑(uik)mxknk=1∑(uik)m,i=1,2,...,c.

The steps of fuzzy C-means clustering are different from the original data standardization, and the physical meanings are different, so it needs to be normalized to obtain the fuzzy vector. For a given sample data set X=x1, x2...xn, the following processing is performed: translation, standard deviation transformation: xkyxkj-x-jsj, where x-j=(nk=1∑xkj)/n, sj=nk=1 ∑(xkj-x-j)2/n"(k=1,2,...,n;j=1,2,...,p).

Xkj is the characteristic parameter in each sample, p is the number of characteristic parameters, p=6 in this paper; xˉj is the mean of the same eigenvalue, and sj is the variance of the same eigenvalue. Translation, range conversion: sample data is converted to (0, 1) according to the principle of membership. Xkj=x'kj-k=nk=1∨x'kjk=1xkjk=nk=1∨xik-k=nk=1∧xkj.

The parameter setting is determined according to the actual situation or using matlab to find the best classification number C of the sample, given the fuzzy weighted index m; taking the iterative step l=0, randomly obtaining the initial classification matrix U(0), stepwise iteration, l=0 , 1, 2,. . . Set the maximum number of iterations T; set an arbitrarily small termination iteration error ε; the iteration start count value is t=0.

Calculate the cluster center vector of the initial classification by calculating the cluster center: V=(V1(l), V2(l),...,Vc(l)), Vi(l)=nk=1∑(uik(l))mxknk =1∑(uik(l))m. Calculate the fuzzy clustering correct rate fuzzy clustering correct rate formula: T=ncn, nc is the correct number of instances; n is the total number of instances. Taking q(n≥q≥1) parameters from any of the n characteristic parameters constitutes the vector xk1, xk2...

Diagnostic examples From August 2002 to July 2005, the valve vibration signals of a unit of natural gas reciprocating compressor were collected periodically, and 16 sets of data were selected as sample data. Firstly, the vibration signals collected in the field are analyzed in time domain and frequency domain, and the sensitive parameters of the fault are selected from the time-frequency domain parameters: frequency domain amplitude maxima, frequency domain mean, time domain dynamic index: peak-to-peak, The absolute value, the effective value, and the variance value are used as the characteristic parameters. The number of classifications is 4, and the fault samples are normalized and then subjected to fuzzy clustering. The fuzzy clustering is used to judge the faults of each fault sample. This shows that based on Table 2 fuzzy clustering grading (a) actual grading (b) situation x2x3x4x5x6x7x8x9x10x11x12x13x14x15x16 zhongliangyouzhong difference good good difference excellent x1 (a) excellent (b) excellent data mining fuzzy clustering method for diagnosis The reciprocating compressor valve failure is effective and accurate.