CN1751857B - Automatic optimizing design method for constant force griding wave filter - Google Patents

Abstract

An automatic optimizing design method for constant-force grinding filter is disclosed. The output signal of filter is y(i)=f(omega1, omega2, x(i)), where omega1 and omega2 are respectively the upper and lower boundary frequencies of pass band and X(i) is input signal. Its algorithm steps includes a), randomly generating initial chromosome (omega 1 omega 2), b), evaluating the chromosome group according to the equation of EFF, c), transfer to f) if the terminating condition is met, d), generating new generation chromosome according to genetic operator, e), returning back to b), and f), stoppingcalculation.

Description

The automatic optimizing design method of constant force griding wave filter

Technical field

The present invention relates to precise machine machining, particularly relate to a kind of technology of automatic optimizing design method of constant force griding wave filter.

Background technology

Constant force griding is a kind of advanced person's a camber grinding method, and it guarantee the even contact of emery wheel and workpiece, thereby grinding goes out high-precision curve by the power of control grinding process.Because the on-line measurement of grinding force is the thing of a difficulty, can adopt the monitoring of acoustic emission signal, judge the contact condition of emery wheel and workpiece.In the grinding condition monitoring, acoustic emission had become a kind of method of extensive employing in recent years.Acoustic emission in the working angles is one of important cutting phenomenon, so-called acoustic emission phenomenon be solid material owing to structural change causes the elastic wave that the quick release of strain energy produces, be called for short AE (Acoustic Emission).In grinding, when grinding was in a kind of stable state, the AE signal did not change, and had only when the grinding state changes, and acoustic emission signal just changes thereupon.In order to differentiate the contact condition of emery wheel and workpiece, generally need at first carry out filtering by wave filter to measuring-signal, calculate the characteristic parameter of reflection grinding state then, and according to certain rule judgment grinding state, as shown in Figure 1.Wherein, Filter Design plays an important role for signal analysis.

Yet, because the radio-frequency component that measured AE signal is comprised, not only relevant with the amount of feeding of emery wheel, also relevant with the size of shape, size and the emery wheel of workpiece, therefore, can only be in wave filter design at a kind of situation, promptly specific a certain part and emery wheel research experiment, determine then filter form (as in low pass filter, high-pass filter and the bandpass filter any) and wave filter by frequency.It is very difficult that this situation is used for the user, can not require the user to each part of being processed all by the research and design wave filter.Therefore, limited the practical application of this technology.

In order to realize constant force griding, at first need to carry out the identification of grinding state, promptly accurately judge emery wheel feed amount whether in predetermined value, as 1 μ m, and realize the control of constant force griding on this basis.The identification of grinding state mainly depends on the calculating of characteristic parameter, and before calculation of characteristic parameters, need carry out the preliminary treatment of measuring-signal, promptly carry out Filter Design, determine the form and the parameter of wave filter, these parameters often need to determine by experience by after a large amount of tests, and change with variations such as cutting material, workpiece sizes, for the practical application of this technology has brought very big difficulty.

Summary of the invention

At the defective that exists in the above-mentioned prior art, technical problem to be solved by this invention provides a kind of based on the biological heredity algorithm, can finish the automatic optimizing design method that the relevant parameter of wave filter is carried out the constant force griding wave filter of characteristic parameter validity evaluation of indexes automatically.

In order to solve the problems of the technologies described above, the automatic optimizing design method of a kind of constant force griding wave filter provided by the invention, the signal output y (i) of its median filter is expressed as:

y(i)＝f(δ ₁，δ ₂，ω ₁，ω ₂，ω _st1，ω _st2，x(i))

δ in the formula ₁, δ ₂Be respectively the maximum attenuation of passband permission and the minimal attenuation that stopband should reach; ω ₁, ω ₂Be respectively the following coboundary frequency of passband; ω _St1, ω _St2Be respectively the following coboundary frequency of stopband, x (i) is for measuring input signal.Based on the biological heredity algorithm to ω ₁, ω ₂Carry out Combinatorial Optimization and calculate, setting the chromosome form is (ω ₁ω ₂), consider actual frequency analysis scope, determine ω ₁, ω ₂Span be (0, ω _Max); Its specific algorithm step is as follows:

1) producing sample size randomly in parameter area is 100 initial chromosome (ω ₁ω ₂);

2) by the formula of characteristic parameter validity evaluation index EFF chromosome population is estimated; The formula of described characteristic parameter validity evaluation index EFF is:

$\mathrm{EFF}=\frac{D_{22}}{D_{21}+D_{23}}$

D in the formula ₂₂Under the predetermined grinding and feeding amount of expression, the distance between the calculation of characteristic parameters value; D ₂₁The predetermined grinding and feeding amount (as 2 μ m) of expression down the calculation of characteristic parameters value with less than the distance between the calculation of characteristic parameters value under the predetermined grinding and feeding amount (as 1 μ m); D ₂₃The predetermined grinding and feeding amount (as 2 μ m) of expression down the calculation of characteristic parameters value with greater than the distance between the calculation of characteristic parameters value under the predetermined grinding and feeding amount (as 3 μ m); EFF is more little, and characteristic parameter is effective more;

3) as satisfying termination condition, go to 6);

4) produce chromosome of new generation according to genetic operator; Described genetic operator is:

Duplicate: in each generation, duplicated several samples of previous generation optimum;

Intersect:

Variation: choose a chromosome (ω randomly ₁ω ₂), and make wherein 1 parameter (ω randomly ₁Or ω ₂) in its span, increase randomly or reduce 5%;

5) be back to 2);

6) stop to calculate.

Signal output y (i)=f (δ of described wave filter ₁, δ ₂, ω ₁, ω ₂, ω _St1, ω _St2, x (i)); Performance requirement according to wave filter in the formula is determined δ ₁, δ ₂Value and ω _St1, ω _St2With ω ₁, ω ₂Relation, the signal of described wave filter output y (i) can be expressed as:

y(i)＝f(ω ₁，ω ₂，x(i))。

Utilize the automatic optimizing design method of constant force griding wave filter provided by the invention,, can determine the optimized parameter ω of wave filter design by above computational process ₁, ω ₂Fig. 2 shows ω ₁, ω ₂Under the various combination situation, the not same-action of wave filter.Work as ω ₁Near 0 o'clock, wave filter was a low pass filter; Work as ω ₂Near ω _MaxThe time, wave filter is a high-pass filter; Other situation wave filters are a bandpass filter.Therefore, method of the present invention can be determined optimum wave filter design form and parameter.

Description of drawings

Fig. 1 is an embodiment of the invention grinding condition discriminating signal processing procedure schematic block diagram;

Fig. 2 is ω in the embodiment of the invention ₁, ω ₂Curve map under the various combination situation.

The specific embodiment

Below in conjunction with description of drawings embodiments of the invention are described in further detail, but present embodiment is not limited to the present invention, every employing similarity method of the present invention and similar variation thereof all should be listed protection scope of the present invention in.

The automatic optimizing design method of the embodiment of the invention provided a kind of wave filter based on genetic algorithm.

For the time domain measurement signal, generally need be by low pass, band is logical or high pass is carried out preliminary treatment to test signal, owing to, therefore the signal output y (i) of wave filter can be expressed as by changing the effect that filter parameter can reach low pass and high-pass filter with bandpass filter:

y(i)＝f(δ ₁，δ ₂，ω ₁，ω ₂，ω _st1，ω _st2，x(i))

δ in the formula ₁, δ ₂Be respectively the maximum attenuation of passband permission and the minimal attenuation that stopband should reach; ω ₁, ω ₂Be respectively the following coboundary frequency of passband; ω _St1, ω _St2Be respectively the following coboundary frequency of stopband, x (i) is for measuring input signal.Performance requirement according to wave filter can be determined δ ₁, δ ₂Value and ω _St1, ω _St2With ω ₁, ω ₂Relation, the argumentation of relevant digital filter design theory in the concrete visible prior art.Therefore the signal of wave filter output y (i) can be expressed as:

y(i)＝f(ω ₁，ω ₂，x(i))

ω ₁, ω ₂Determine to make that the characteristic parameter validity of being calculated by y (i) is best.Effectively characteristic parameter should make class interior (under the same grinding and feeding amount, the calculated value of characteristic parameter) distance is the smaller the better, and (under the different grinding and feeding amounts, the calculated value of characteristic parameter) distance is the bigger the better between class, so the validity evaluation index of defined feature parameter is

$\mathrm{EFF}=\frac{D_{22}}{D_{21}+D_{23}}$

D in the formula ₂₂Under the predetermined grinding and feeding amount of expression, the distance between the calculation of characteristic parameters value; D ₂₁The predetermined grinding and feeding amount (as 2 μ m) of expression down the calculation of characteristic parameters value with less than the distance between the calculation of characteristic parameters value under the predetermined grinding and feeding amount (as 1 μ m); D ₂₃The predetermined grinding and feeding amount (as 2 μ m) of expression down the calculation of characteristic parameters value with greater than the distance between the calculation of characteristic parameters value under the predetermined grinding and feeding amount (as 3 μ m); EFF is more little, and characteristic parameter is effective more.Can carry out ω according to the biological heredity algorithm ₁, ω ₂Combinatorial Optimization calculate.

Setting the chromosome form is (ω ₁ω ₂), consider actual frequency analysis scope, determine ω ₁, ω ₂Span be (0, ω _Max), genetic operator is

Duplicate: in each generation, duplicated several samples of previous generation optimum;

Intersect:

Variation: choose a chromosome (ω randomly ₁ω ₂), and make wherein 1 parameter (ω randomly ₁Or ω ₂) in its span, increase randomly or reduce 5%.

Its specific algorithm step is as follows:

1) producing sample size randomly in parameter area is 100 initial chromosome (ω ₁ω ₂).

2) by the formula of characteristic parameter validity index EFF chromosome population is estimated.

3) as satisfying termination condition, go to 6).

4) produce chromosome of new generation according to above-mentioned genetic operator.

5) be back to 2).

6) stop to calculate.

Above interative computation process when reaching predetermined number of iterations till.Evaluation function in the algorithm is mainly used in to be estimated each chromosome in the current generation, in the next generation, remove the chromosome of several low performances by estimating, keep some high performance chromosomes, and by the more additional new chromosomes of genetic operator, obtain very outstanding colony at last, satisfy the requirement of problem solving.

Claims (1)

1. the automatic optimizing design method of a constant force griding wave filter, the signal output y (i) of its median filter is expressed as:

y(i)＝f(δ ₁，δ ₂，ω ₁，ω ₂，ω _st1，ω _st2，x(i))

δ in the formula ₁, δ ₂Be respectively the maximum attenuation of passband permission and the minimal attenuation that stopband should reach; ω ₁, ω ₂Be respectively the following coboundary frequency of passband; ω _St, ω _St2Be respectively the following coboundary frequency of stopband, x (i) is for measuring input signal; Setting the chromosome form is (ω ₁ω ₂), and definite ω ₁, ω ₂Span be (0, ω _Max); Its specific algorithm step is as follows:

1) producing sample size randomly in parameter area is 100 initial chromosome (ω ₁ω ₂);

2) by the formula of characteristic parameter validity evaluation index EFF chromosome population is estimated; The formula of described characteristic parameter validity evaluation index EFF is:

$\mathrm{EFF}=\frac{D_{22}}{D_{21}+D_{23}}$

D in the formula ₂₂Under the predetermined grinding and feeding amount of expression, the distance between the calculation of characteristic parameters value; D ₂₁Under the predetermined grinding and feeding amount of expression the calculation of characteristic parameters value with less than being scheduled to the distance between the calculation of characteristic parameters value under the grinding and feeding amount; D ₂₃Under the predetermined grinding and feeding amount of expression the calculation of characteristic parameters value with greater than being scheduled to the distance between the calculation of characteristic parameters value under the grinding and feeding amount;

3) as satisfying termination condition, go to 6);

4) produce chromosome of new generation according to genetic operator; Described genetic operator is:

Duplicate: in each generation, duplicated several samples of previous generation optimum;

Intersect:

Variation: choose a chromosome (ω randomly ₁ω ₂), and make wherein 1 parameter ω randomly ₁Or ω ₂In its span, increase randomly or reduce 5%;

5) be back to 2);

6) stop to calculate;

Signal output y (i)=f (δ of described wave filter ₁, δ ₂, ω ₁, ω ₂, ω _St1, ω _St2, x (i)); Performance requirement according to wave filter in the formula is determined δ ₁, δ ₂Value and ω _St1, ω _St2With ω ₁, ω ₂Relation, the signal of described wave filter output y (i) can be expressed as:

y(i)＝f(ω ₁，ω ₂，x(i))；

By above computational process, can determine the optimized parameter ω of wave filter design ₁, ω ₂Thereby, can determine optimum wave filter design form and parameter.

Images