CN103072302A - Braking curve self-learning method for numerical control press - Google Patents

Abstract

The invention discloses a braking curve self-learning method for a numerical control press. The method comprises the following steps of: (1) setting n different frequencies corresponding to n different speeds of the press, braking the press at the end of an action cycle of the press, and recording a speed before the braking of each time and an overshoot angle after the braking after the press is stopped to obtain n pieces of point data, wherein n is a natural number more than or equal to 5; and (2) fitting a quadric curve according to the n pieces of recorded point data by adopting a least square algorithm, and storing three coefficients of the quadric curve to obtain a braking curve. The braking curve self-learning method for the numerical control press is convenient to operate and high in accuracy.

Description

A kind of brake cruve self-learning method of numerical control press

Technical field

The present invention relates to a kind of brake cruve control method of numerical control press.

Background technology

At present, the fast development of domestic numerical control press industry, the control system demand that is applied to punch press is increased sharply; The punch brake curve refers to behind the punch brake because the relation of the angle of rotary inertia overshoot and the front velocity of rotation of brake utilizes this curve can guarantee to stop near top dead-centre behind the punch brake.The punch press control system of current use adopts mostly take PLC as main control mode, is not suitable for the occasion that some need complex calculation, causes the undesirable of punch brake curve precision.And the brake cruve learning automaton degree that adopts is not high enough, and each study needs artificial several different speed that arrange to record the parking angle.And the brake cruve parameter of every machine is different, so every machine all needs artificial study.Bring inconvenience to a certain extent to operating personnel.

Summary of the invention

For the lower deficiency of troublesome poeration, precision of the brake cruve learning method that overcomes existing numerical control press, the invention provides a kind of brake cruve self-learning method of easy to operate, numerical control press that precision is higher.

The technical solution adopted for the present invention to solve the technical problems is:

A kind of brake cruve self-learning method of numerical control press said method comprising the steps of: (1) arranges n different frequency, corresponding punch press n different speed, and n is natural number, and n 〉=5; The punch press action cycle is finished rear braking, waits punch press to stop the angle of overshoot after speed before the each brake of rear record and the brake, namely obtains n and puts data;

(2) n some data according to record simulate a conic section, and described approximating method adopts least-squares algorithm, stores three coefficients of this conic section, obtains brake cruve.

Further, in the described step (2), three coefficients are shown on the touch-screen for artificial fine setting.

Beneficial effect of the present invention is mainly manifested in: easy to operate, precision is higher.

Description of drawings

Fig. 1 is punch brake curve self study program flow diagram.

Fig. 2 is least square fitting conic section program flow diagram.

The specific embodiment

The invention will be further described below in conjunction with accompanying drawing.

See figures.1.and.2, a kind of brake cruve self-learning method of numerical control press said method comprising the steps of: (1) arranges n different frequency, corresponding punch press n different speed, and n is natural number, and n 〉=5; The punch press action cycle is finished rear braking, waits punch press to stop the angle of overshoot after speed before the each brake of rear record and the brake, namely obtains n and puts data;

(2) n some data according to record simulate a conic section, and described approximating method adopts least-squares algorithm, stores three coefficients of this conic section, obtains brake cruve.

Further, in the described step (2), three coefficients are shown on the touch-screen for artificial fine setting.

Punch brake curve self-learning method is mainly realized by master control borad, is started by the touchscreen button input.Its flow process as shown in Figure 1; at first master control borad must be placed under the adjusting microinching pattern; then the master control borad program can judge whether receive the curve learning signal in the situation of shutting down; in case receive the curve learning signal; just begin by 485 communications setting frequency converter frequencies, thereby the motor initial velocity is set, after the wait motor speed is steady; the auto-closing clutch starts 10 all after date cut-off clutches of punch press action, and the overshoot angle after recording punch press velocity of rotation before and stopping.Next judge whether to have learnt the point of all settings, do not increase 10r/min if just continue to arrange speed, again record next curve point, until learnt the curve point of setting.Just begin afterwards to adopt least-squares algorithm that 10 curve point of before study are fitted to a conic section.The later on each orderly shutdown of punch press just can adopt this curve to calculate the angle of braking in advance, and slider of punch is stopped near top dead-centre.

Least square fitting conic section process is as follows: at first go out the augmented matrix B of normal equation group according to several curve point coordinate Calculation of front learning records, then matrix B is changed into upper triangular matrix, calculate at last three coefficients of conic section.Because the internal memory of this MCU is larger, therefore can directly adopt the matrix operation of least square method to find the solution, compare the direct employing method of undetermined coefficients and find the solution and greatly reduced amount of calculation.

In the present embodiment, the hardware platform that the realization of punch brake curve self study is adopted is based on the embedded scheme of the stm32 microcontroller (MCU) of arm-cortex-m3.Whole punch press control system is by forming take stm32f103rb as main touch-screen HMI terminal with take stm32f103ze as main master control borad, between the two by 232 serial communication synchrodatas.The study of brake cruve at first will place master control borad the adjusting microinching pattern, then after pressing the curve learn button on the touch-screen and inputting password, the touch-screen display board can send curve study order to master control borad by 232 serial ports, master control borad is received after the information can be automatically to communicate by letter with frequency converter 485 and is set gradually 10 different frequencies, 10 different speed of corresponding punch press, move several cycle rear brakings, then the angle (displacement) of overshoot after waiting punch press to stop the front speed of the each brake of rear record and brake, these several points according to record simulate a conic section at last, and store three coefficients of this conic section, simultaneously these three coefficients are shown on the touch-screen for artificial fine setting.Whole process need not manual intervention, and the match of brake cruve also abandoned the segmented fitting method that former PLC control system adopts, and has adopted the conic fitting method of least-squares algorithm, has improved precision and the adaptability of curve study.This more meets behind the punch brake conic section physical characteristic of speed before the overshoot angle and brake.

The least square fitting conic section is described as follows: establishing the front speed of each brake is x, and the overshoot angle after killing is y, and it satisfies conic section as can be known according to its physical characteristic, so can establish:

f(x)=a ₀+a ₁*x+a ₂*x ² （1）

y=f(x) （2）

Several curve point coordinate (x with the front learning records ₁, y ₁), (x ₂, y ₂) ..., (x ₁₀, y ₁₀) coordinate is distinguished substitution (1), (2) then can get according to least square method:

$g\left(x\right)=\underset{i=1}{\overset{10}{\mathrm{\Σ}}}{(y_i-f\left(x_i\right))}^2---\left(3\right)$

G (x) three coefficient a when obtaining minimum of a value in (3) formula ₀, a ₁, a ₂Near actual value.Therefore with g (x) to a ₀, a ₁, a ₂Partial derivative equal respectively 0:

$\frac{\∂g\left(x\right)}{\∂a_j}=2\underset{i=1}{\overset{10}{\mathrm{\Σ}}}[y_i-f\left(x_i\right)]{x_i}^j=2\underset{i=0}{\overset{10}{\mathrm{\Σ}}}[y_i-\underset{k=0}{\overset{2}{\mathrm{\Σ}}}{a}_k{x}_i^k]x_i^j=0---\left(4\right)$

Namely $\underset{k=0}{\overset{2}{\mathrm{\Σ}}}{a}_k\left(\underset{i=1}{\overset{10}{\mathrm{\Σ}}}{x}_i^{k+j}\right)=\underset{i=1}{\stackrel{10}{\mathrm{\Σ}}}{y}_i{x}_i^j$ J=0 wherein, 1,2.（5）

By (5) equation group of can doing in the proper way:

$\left\{\begin{array}{c}10a_0+a_1\underset{i=1}{\overset{10}{\mathrm{\Σ}}}{x}_i+a_2\underset{i=1}{\overset{10}{\mathrm{\Σ}}}{x}_i^2=\underset{i=1}{\overset{10}{\mathrm{\Σ}}}{y}_i\\ a_0\underset{i=1}{\overset{10}{\mathrm{\Σ}}}{x}_i+a_1\underset{i=1}{\overset{10}{\mathrm{\Σ}}}{x}_i^2+a_2\underset{i=1}{\overset{10}{\mathrm{\Σ}}}{x}_i^3=\underset{i=1}{\overset{10}{\mathrm{\Σ}}}{y}_i{x}_i\\ a_0\underset{i=1}{\overset{10}{\mathrm{\Σ}}}{x}_i^2+a_1\underset{i=1}{\overset{10}{\mathrm{\Σ}}}{x}_i^3+a_2\underset{i=1}{\overset{10}{\mathrm{\Σ}}}{x}_i^4=\underset{i=1}{\overset{10}{\mathrm{\Σ}}}{y}_i{x}_i^2\end{array}\right.---\left(6\right)$

Writing the established law matrix form is:

$\left(\begin{array}{ccc}10& \underset{i=1}{\overset{10}{\mathrm{\Σ}}}{x}_i& \underset{i=1}{\overset{10}{\mathrm{\Σ}}}{x}_i^2\\ \underset{i=1}{\overset{10}{\mathrm{\Σ}}}{x}_i& \underset{i=1}{\overset{10}{\mathrm{\Σ}}}{x}_i^2& \underset{i=1}{\overset{10}{\mathrm{\Σ}}}{x}_i^3\\ \underset{i=1}{\overset{10}{\mathrm{\Σ}}}{x}_i^2& \underset{i=1}{\overset{10}{\mathrm{\Σ}}}{x}_i^3& \underset{i=1}{\overset{10}{\mathrm{\Σ}}}{x}_i^4\end{array}\right)\·\left(\begin{array}{c}{a}_0\\ a_1\\ a_2\end{array}\right)=\left(\begin{array}{c}\underset{i=1}{\overset{100}{\mathrm{\Σ}}}{y}_i\\ \underset{i=1}{\overset{10}{\mathrm{\Σ}}}{x}_i{y}_i\\ \underset{i=1}{\overset{10}{\mathrm{\Σ}}}{x}_i^2{y}_i\end{array}\right)---\left(7\right)$

Just can draw augmented matrix B according to (7):

$\left(\begin{array}{cccc}10& \underset{i=1}{\overset{10}{\mathrm{\Σ}}}{x}_i& \underset{i=1}{\overset{10}{\mathrm{\Σ}}}{x}_i^2& \underset{i=1}{\overset{10}{\mathrm{\Σ}}}{y}_i\\ \underset{i=1}{\overset{10}{\mathrm{\Σ}}}{x}_i& \underset{i=1}{\overset{10}{\mathrm{\Σ}}}{x}_i^2& \underset{i=1}{\overset{10}{\mathrm{\Σ}}}{x}_i^3& \underset{i=1}{\overset{10}{\mathrm{\Σ}}}{x}_i{y}_i\\ \underset{i=1}{\overset{10}{\mathrm{\Σ}}}{x}_i^2& \underset{i=1}{\overset{10}{\mathrm{\Σ}}}{x}_i^3& \underset{i=1}{\overset{10}{\mathrm{\Σ}}}{x}_i^4& \underset{i=1}{\overset{10}{\mathrm{\Σ}}}{x}_i^2{y}_i\end{array}\right)---\left(8\right)$

Just can be solved three coefficients of conic section through matrix operation by matrix B.

Claims (2)

1. the brake cruve self-learning method of a numerical control press is characterized in that: said method comprising the steps of: (1) arranges n different frequency, corresponding punch press n different speed, and n is natural number, and n 〉=5; The punch press action cycle is finished rear braking, waits punch press to stop the angle of overshoot after speed before the each brake of rear record and the brake, namely obtains n and puts data;

(2) n some data according to record simulate a conic section, and described approximating method adopts least-squares algorithm, stores three coefficients of this conic section, obtains brake cruve.

2. the brake cruve self-learning method of a kind of numerical control press as claimed in claim 1 is characterized in that: in the described step (2), three coefficients are shown on the touch-screen for artificial fine setting.

Images