WO2023051072A1 - Intelligent debugging method and system based on optimization compensation - Google Patents

Abstract

An intelligent debugging method and system based on optimization compensation, which method and system are used for performing intelligent debugging on at least one numerical control machine (10). The method comprises the following steps: S1, acquiring a measurement model parameter of a numerical control machine (10), and receiving actual measurement data of a machined part that is collected by the numerical control machine (10); S2, according to the measurement model parameter, converting, into measurement data in a standard format, the actual measurement data of the machined part that is collected by the numerical control machine (10); S3, according to the measurement data in the standard format and the measurement model parameter, establishing a weighted optimization problem model based on a trust region, and solving the optimization problem model to obtain an optimization compensation value of the numerical control machine (10); and S4, returning the optimization compensation value to the numerical control machine (10), such that the numerical control machine (10) performs debugging according to the optimization compensation value. By means of the intelligent debugging method and system based on optimization compensation, the problem in the prior art that is caused by a possible correlation between sizes during "debugging" is solved.

Description

Intelligent adjustment method and system based on optimal compensation technical field

The invention relates to the technical field of numerical control machining, in particular to an intelligent adjustment method and system based on optimal compensation.

Background technique

In the CNC metal processing industry, the final processing of a part according to the design drawings needs to be decomposed into multiple stages: design drawings - DFM process dismantling - NPI processing verification - mass production. The DFM process dismantling includes dismantling a part into multiple processing procedures, the processing description, equipment requirements, process drawings, tool fixture information, SIP inspection standards, and program overview of each process; when entering the NPI processing verification stage, it is necessary to select Fix the machine table, fixtures, tools, and carry out production verification. "Machine adjustment" is the core link of production verification. What it needs to do is to adjust the tool compensation value and coordinate system compensation value used in part processing to make all dimensions meet the quality standards. The success of "tuning" is affected by various upstream and downstream conditions, including: the rationality of process dismantling, the practicability of quality control, the performance stability of equipment used in production, and the assembly of fixtures. The combination of standardization, tool selection and method of cutting procedures, etc., all affect the tuning results. Therefore, "tuning" is the fundamental determinant of the completion speed of the entire production verification stage, and the verification of the results of each related work process is reflected by the results of "tuning". The faster the "tuning" is passed, the sooner the company can receive mass production orders and complete product delivery.

In the existing CNC metal processing field, the basic process of "adjusting the machine" is: when each machine produces and processes the first piece of material, the production technicians will rely on visual observation and work experience to manually adjust the tool compensation and coordinates used for processing The compensation value is used to produce finished parts, and then the parts are sent to the offline equipment for inspection and measurement to see if all dimensions meet the standards. If the standard is met, the adjustment machine will pass. If it does not meet the standard, change the material, process and adjust the machine again, and send it for inspection until all dimensions of a finished product are qualified. It is generally necessary to repeatedly adjust the machine to meet the product size requirements. Since each size may be related, one knife or one coordinate system will participate in the processing of multiple sizes. As a result, the size of a problem may be adjusted to meet the standard, while another related size has a problem. Therefore, in the existing "machine adjustment" process, the necessary condition is "sufficient quantity and rich technical experience" production technicians. Only by ensuring this condition can the production verification phase be completed as scheduled, and the technical verification can be passed as scheduled, thereby ensuring the smooth acquisition of mass production orders. The reality is that it is becoming more and more difficult to recruit production technicians, and the corresponding labor costs are also increasing, which has become a major pain point for manufacturing companies.

If the enterprise can ensure that the production technicians are in place and carry out the verification of machine adjustment normally, the next thing they will face is the difficulty of "impeded efficiency". The time-consuming process of "machine adjustment" is not only the time for adjusting the tool compensation and coordinate system compensation for reprocessing, but also in the inspection process of the product after each machine adjustment and production. At present, the inspection of the adjustment results of the factory is mostly carried out outside the machine, that is, after a piece of parts is produced, it must be taken to the laboratory, using professional testing equipment, and outputting a test report. The ratio gap between testing equipment and production equipment is huge, reaching 1.5:100. Like this, naturally produced a large amount of queuing to wait for detection time, a piece of material just may need to wait several hours, even longer. In this way, it may take several days for a CNC processing equipment to be adjusted, and it will take two months to complete the adjustment of hundreds of equipment. The time spent on "adjustment" is also the loss of available time for factory equipment. Improving the efficiency of "adjustment" and speeding up the production verification stage to the greatest extent can release the available time of equipment to the greatest extent, and then convert it into production capacity.

In addition to the personnel and efficiency problems faced by the lowering machine, the problem of "material waste" also occurred. The parts produced by the failure of machine adjustment often cannot be processed twice, forming a piece of waste. Then the higher the one-time completion rate of machine adjustment, the less material loss. The resulting yield data will be better. At the same time, in addition to the need to adjust the machine during the production capacity verification stage mentioned above, it also needs to be adjusted during the mass production process. For example, every day, when each machine is making the first piece of material, when the clamping fixture inside the machine is updated, and after the tool is changed, it is necessary to carry out a machine adjustment test on the subsequent material to ensure the quantity Yield rate of production. Therefore, the scene of "adjustment" runs through every link of CNC metal processing, and solving the problem of "adjustment" will bring about a huge increase in production capacity.

The disclosure of the above background technical content is only used to assist in understanding the concept and technical solution of the present invention, and it does not necessarily belong to the prior art of this patent application. If there is no clear evidence that the above content has been disclosed on the filing date of this patent application In some cases, the above background technology should not be used to evaluate the novelty and inventiveness of this application.

Contents of the invention

In order to solve the problems caused by the possible correlation between the various dimensions during the above "adjustment", the present invention proposes an intelligent adjustment method and system based on optimal compensation.

In order to achieve the above object, the present invention adopts the following technical solutions:

One embodiment of the present invention discloses an intelligent machine tuning method based on optimal compensation, which is used for intelligent machine tuning of at least one numerical control machine, including the following steps:

S1: Obtain the measurement model parameters of the CNC machine, and receive the actual measurement data of the processed parts collected by the CNC machine;

S2: Convert the actual measurement data of the processed parts collected by the CNC machine into measurement data in a standard format according to the measurement model parameters;

S3: Establish a weighted optimization problem model based on the trust region according to the measurement data and measurement model parameters in a standard format, and solve the optimization problem model to obtain the optimal compensation value of the numerical control machine;

S4: Send the optimized compensation value back to the numerical control machine, so that the numerical control machine performs machine adjustment according to the optimized compensation value.

Preferably, the measurement model parameters obtained in step S1 include the type, standard value and tolerance standard of each dimension to be measured of the machined part of the CNC machine, and the tool compensation coefficient of the CNC machine.

Preferably, step S1 further includes: correcting the actual measurement data of the processed part collected by the numerical control machine by using the point deformation of the standard block.

Preferably, using standard block point deformation to correct the actual measurement data collected by the CNC machine specifically includes: providing standard blocks corresponding to the dimensions of the processed parts to be measured, and each standard block is provided with a plurality of Calibration point, obtain the reference value of each calibration point of the standard block corresponding to each size to be measured; when receiving the actual measurement data of the processed parts collected by the numerical control machine, also receive each The actual measurement data of the processed parts collected by the numerical control machine is corrected according to the reference value and the calibration value of each calibration point of the standard block corresponding to each size to be measured.

Preferably, receiving the actual measurement data of the processed parts collected by the numerical control machine in step S1 specifically includes: receiving the actual measurement data of the processed parts collected by the numerical control machine by establishing a handshake protocol with the numerical control machine.

Preferably, receiving the actual measurement data of the processed parts collected by the numerical control machine by establishing a handshake agreement with the numerical control machine specifically includes:

A1: Establish a long link with the CNC machine to notify the CNC machine to enter the collection state when receiving data;

A2: Receive the feature number transmitted by the CNC machine in the form of macro variables and a corresponding point measurement data collected;

A3: Reset the macro variable, and judge whether the CNC machine has collected all the points, if yes, complete the reception, if not, notify the CNC machine to enter the measurement and reporting of the next point, and return Step A2.

Preferably, step S2 further includes: judging according to the measurement model parameters whether the obtained measurement data in a standard format is up to standard, and if not, sending an alarm signal to the numerical control machine, so that the numerical control machine suspends processing Component.

Preferably, step S3 specifically includes: the weighted optimization problem model based on the trust region established according to the measurement data in the standard format and the measurement model parameters is expressed as follows:

s.t.lsp≤x≤usp

Ax=b

And using the trust region reflection method to solve the above-mentioned optimization problem model to obtain the optimal compensation value of the numerical control machine;

Among them, A=[z,I] ^T ∈ ^{R n×(n+k)} , x=[X,ε]∈R ^n+k , b=CS∈R ⁿ , C∈R ⁿ is the measurement data in the standard format , S∈R ⁿ is the standard value of each dimension of the processed part included in the measurement model parameters, z∈R ^k×n is the tool compensation coefficient of the CNC machine included in the measurement model parameters, ε∈R ⁿ is the The tolerance of the CNC machine to be solved, X∈R ^k is the compensation value of the CNC machine to be solved, I is the unit matrix, E∈R ^n×(n+k) is the weight, lsp, usp∈R ^n+k are respectively are the upper and lower bounds of the unknown quantity x.

Preferably, using the trust region reflection method to solve the above optimization problem model to obtain the optimal compensation value of the numerical control machine specifically includes:

Construct function f(x), and calculate X and ε by minimizing f(x) on trust region N through trial steps s, where trust region N is the neighborhood of function f(x) at point x.

Preferably, minimizing f(x) on the trust region N through the trial step s specifically includes:

B1: Construct a two-dimensional trust region subproblem:

Among them, g is the gradient of the function f at the current point x, H is the Hessian matrix, D is the diagonal scaling matrix, s is the trial step, and Δ is the radius of the trust region;

B2: Solve the two-dimensional trust region subproblem to determine the trial step s;

B3: Determine whether f(x+s) is smaller than f(x), if yes, make x=x+s, and return to step B1; if not, keep the current point unchanged, reduce the trust region radius Δ, and return Step B1;

B4: Repeat steps B1-B3 until the two-dimensional trust region subproblem converges, and the solution to the optimization problem f(x) is obtained.

Preferably, when solving the two-dimensional trust region subproblem in step B2, the two-dimensional trust region subproblem is limited to the two-dimensional subspace K for solution, and the two-dimensional subspace K is determined according to the following preconditioned conjugate gradient method: Define the two-dimensional subspace K as a linear space determined by k ₁ and k ₂ , where k ₁ is the direction of the gradient g, and k ₂ satisfies H·k ₂ =-g or

conditions of.

Preferably, after using the trust region reflection method to solve the above optimization problem model to obtain the optimal compensation value of the CNC machine, it also includes: the optimal compensation value of the CNC machine obtained according to the optimization problem model and the measurement The model parameters correct the optimal compensation value of the numerical control machine.

Preferably, before sending the optimized compensation value back to the numerically controlled machine in step S4, it is judged whether the optimized compensated value exceeds a preset threshold, and if so, an intervention signal is sent to the numerically controlled machine, so that The numerical control machine suspends processing parts; if not, the optimized compensation value is sent back to the numerical control machine.

Preferably, before sending the optimized compensation value back to the CNC machine in step S4, it is judged whether the optimized compensation value exceeds a preset threshold, and if so, the standard block is used again to collect the data collected by the CNC machine. The actual measurement data of the processed part is corrected; if the threshold is still exceeded after correction, an intervention signal is sent to the CNC machine to make the CNC machine suspend processing the part; if the preset threshold is not exceeded, the optimized The compensation value is sent back to the numerical control machine.

Another embodiment of the present invention discloses an intelligent adjustment system based on optimization compensation, which is used for intelligent adjustment of at least one numerical control machine, and is characterized in that it includes: a processor and a storage medium, and the storage medium A computer program is stored in the computer, and the processor is configured to run the computer program to execute the above-mentioned intelligent tuning method.

Compared with the prior art, the beneficial effect of the present invention lies in: the intelligent adjustment method and system based on optimization compensation proposed by the present invention, wherein by receiving the actual measurement data and measurement model parameters of the machined parts of the CNC machine, adopting the method based on trust The weighted optimization problem model of the domain solves the optimal compensation value of the CNC machine to guide the machine adjustment of the CNC machine, making the machine adjustment a standardized process. Among them, the tool compensation or coordinate compensation value can no longer only rely on empirical calculations. The entire process of machine adjustment and compensation does not require professional manual participation, and only ordinary operators need to perform processing, loading, unloading and processing operations. As a result, the factory has completely solved the problem of difficulty in recruiting workers for biotechnology and high cost. Moreover, the algorithm and measurement method used in the intelligent tuning method make the accuracy and precision of the tuning compensation parameters reach the best, and the tuning effect is beyond the reach of excellent biotechnology. At the same time, this kind of accuracy and precision will not be different due to different equipment and environments, and it is extremely stable. In addition, the intelligent adjustment system has increased the efficiency of machine adjustment by more than half; in the past, machine adjustment and material preparation had to be repeated 6 times or more. After using the machine adjustment method and system, it can be done in one go. In the past, with manual participation, repeated adjustments needed to go back and forth to the laboratory or offline testing equipment for judgment. The time-consuming costs included "offline testing time", "manual judgment time", "repeated processing time", "waiting for testing equipment Time (considering that the number of testing equipment is far less than that of production equipment, about 1.5:100, the processed parts need to be tested offline, and you need to wait in line for the idle time of the equipment)", while the machine adjustment time is only "two batches of material processing Time" and "Run time of one internal measurement program". In the climbing stage of new product verification, using the intelligent tuning method and system can reduce the original tuning time by half. At the same time, since most of the size measurement is transferred to the CNC machine, the problem is exposed and solved in the machine measurement in advance, which also greatly reduces the pressure on the demand for laboratory measuring instruments.

In a further solution, by receiving the actual measurement data and measurement model parameters of the machined parts of the CNC machine, the detection of the adjustment result is carried out in the machine, and the actual data of the machined parts collected by the CNC machine is measured with a standard block. The measurement data is corrected, saving the adjustment time.

In a further solution, the actual measurement data of the processed parts collected by the CNC machine is received by establishing a handshake protocol based on system macro variables with the CNC machine, so that the measurement can be carried out inside the machine, saving the adjustment time. By establishing the above-mentioned handshake protocol, the normal processing of the CNC machine can not be affected, the handshake efficiency is high, the macro variables occupying the CNC machine are small, the number of handshakes is small, but the data interaction can be guaranteed to be stable, and there will be no data loss or wrong collection and reception.

Description of drawings

Fig. 1 is the flow chart of the intelligent adjustment method based on optimal compensation of the preferred embodiment of the present invention;

Fig. 2 is the schematic diagram of the standard block of a specific embodiment of the present invention;

Fig. 3 is the flow chart of the intelligent tuning machine method of the specific embodiment of the present invention;

Fig. 4 is the frame diagram of the intelligent tuning machine system of the specific embodiment of the present invention;

FIG. 5 is a schematic diagram of correcting the result value of workpiece measurement.

Detailed ways

The present invention will be further described below with reference to the accompanying drawings and in combination with preferred embodiments.

The preferred embodiment of the present invention discloses an intelligent adjustment method based on optimization compensation, which is used for intelligent adjustment of at least one numerical control machine, so that the adjustment process is smoother, and the adjustment compensation data reaches the target at one time. The target process is: after the machining of the CNC machine is completed, the probe is used to measure the size points in the machine. The system collects the measurement points in real time to calculate each size value of the part. The system page visually displays the size measurement results of the processed parts and automatically calculates all The optimal value of the tool compensation and coordinate system compensation involved in the processing, the compensation value can be written back to the machine with one click of the compensation button, and the CNC machine can be processed and measured again to produce products of various sizes that meet the standards. This adjustment process is not limited by the CNC machine system, and is applicable to all types of CNC machines.

As shown in Figure 1, the intelligent tuning method specifically includes the following steps:

S1: Obtain the measurement model parameters of the CNC machine, and receive the actual measurement data of the processed parts collected by the CNC machine;

Wherein, the acquired measurement model parameters include the type, standard value and tolerance standard of each dimension to be measured of the machined parts of the CNC machine, the tool compensation coefficient of the CNC machine, and the like.

When a part starts to be processed, verified and adjusted, each CNC machine needs to pass the adjustment. That is to ensure that each CNC machine is capable of producing parts that meet the quality requirements, or when the first piece of material is produced every day, machine adjustment verification is also required. The basis of the machine adjustment work is that the process and programming departments have determined the drawing version of the processed parts, formulated the process methods (such as process splitting, processing NC programs, processing tools, processing fixtures, processing equipment performance requirements, etc.), and have determined Quality inspection requirements, such as the size of the inspection, and its tolerance standards. For example, for a length dimension, the standard is 0.83cm. If the upper line is within the range of 0.09cm and the lower line is 0.05cm, the standard is met, that is, the produced size is 0.78cm to 0.92cm, and it is judged as qualified.

Specifically, before starting machine adjustment, the following data content needs to be stored: (1) processed items; (2) processing equipment allocated for the items; (3) parts under the items; (4) each drawing version of the parts; 5) Multiple processes under each drawing version; (6) Each process contains a series of inspection dimensions; (7) Each dimension has a set of tolerance standards, a set of processing tool information, and a set of processing coordinate system information . Both tool and coordinate system information have actual influence information, such as the tool length or tool radius will affect the processing of this dimension, and the influence coefficient will be recorded, that is, the coefficient relationship between tool/coordinate system compensation and size. For different sizes, under different tool cutting methods, the ratio of coefficients affected by different tool compensations. (7) There are multi-point sizes for each size. That is, starting from different points, the measured values are different. Such as thickness size, radius size, etc., there will be multi-point data for display.

The NC machine is equipped with probe hardware and uses the "measurement NC program" to detect a series of point values in the processed parts; through the part drawings, the relationship between these point positions and the dimensions of the parts is obtained, and all points are calculated through the point values. size value. In the measurement process, there are often some error problems: errors caused by thermal deformation, errors caused by burrs, errors caused by deformation, etc. In this embodiment, after receiving the actual measurement data of the processed parts collected by the CNC machine, the point deformation of the standard block is also used to correct the actual measurement data of the processed parts collected by the CNC machine to solve the above error problem.

In a specific embodiment, the schematic diagram of the standard block is shown in Figure 2, there is one standard block in each of the X and Y directions, and the size of the "marking point" of the standard block is measured and calibrated by three coordinates in advance (marked in the figure Reference value). After the standard block is installed, measure the standard block with the built-in probe. After finding the origin, the machine measures the coordinate values (ie, the calibration value) of all "calibration points" in the standard block. In this way, the error value between the in-machine detection value and the three-coordinate detection value of each calibration point can be calculated, as shown in Table 1. For example, taking the length from A1 to A2 as an example, its theoretical diameter is |A2-A1|=24mm. The online measurement coordinate of point A1 is -118.0, and the closest "calibration point" at this position on the standard block is point No. 9 in Table 1, with a value of -117.0 and an error of 0.009. Therefore, take the error value of this calibration point to correct the A1 point, and the corrected A1 coordinate is -118.0-0.009=-118.009. Similarly, the coordinates of point A2 are corrected -142.0-0.011=-142.011. After point A1 and A2 are corrected, the diameter of the circle |A2-A1|=24.002mm. In the subsequent compensation calculation and value acquisition, the coordinate values that have been compensated by the standard block are taken for the next calculation.

Table 1 The reference value, calibration value and corresponding error of each calibration point of the standard block

Wherein, receiving the actual measurement data of the processed parts collected by the CNC machine is specifically: receiving the actual measurement data of the processed parts collected by the CNC machine by establishing an interactive handshake protocol with the CNC machine, further specifically including the following steps:

A1: Establish a long link with the CNC machine to notify the CNC machine to enter the collection state when receiving data;

A2: Receive the feature number transmitted by the CNC machine in the form of macro variables and the corresponding point measurement data collected; that is, every time the CNC machine obtains a measurement point result, the point result and the agreed The feature number of the system is transmitted to the intelligent tuning system in the form of macro variables; among them, five system macro variables are used for information interaction, which are "program name", "size number", "X value", "Y value", "Z value".

A3: Reset the macro variable, and judge whether the CNC machine has collected all the points. If so, complete the reception, then notify the CNC machine to start the measurement and reporting of the next point, and return to step A2. That is, when all the measurement point data are measured, the CNC machine sends an end signal, and the intelligent machine adjustment system obtains the end signal, ends a round of data reception, and reports all the results of the collected point data to the intelligent machine adjustment system After receiving the data, the server will complete the subsequent size calculation, size judgment and compensation calculation.

By establishing the above-mentioned handshake protocol, the normal processing of the CNC machine can not be affected, the handshake efficiency is high, the macro variables occupying the CNC machine are small, the number of handshakes is small, but the data interaction can be guaranteed to be stable, and there will be no data loss or wrong collection and reception.

S2: Convert the actual measurement data of the processed parts collected by the CNC machine into measurement data in a standard format according to the measurement model parameters;

Find the relationship between the measurement point and the detection size in the part drawing, and calculate the result value of the size through the shape and geometry formula. This part of the operation can be completed within the intelligent detection and adjustment system, or it can be completed by relying on the docking of existing detection systems in the market. For example, to calculate the dimensional values of "circularity", "coaxiality", "parallelism", "perpendicularity", "symmetry", etc., the dimensions to be measured can be marked on the drawing, and the system guides the selection of the required calculation. Measure point. In this way, as long as the inspection program measures the (x, y, z) coordinates of these points, the system can use the established geometric calculation model to automatically complete the calculation of the size results and display them in the system.

Wherein, after the actual measurement data is converted into the measurement data in the standard format, it is also judged according to the measurement model parameters whether the obtained measurement data in the standard format is up to the standard, and if it is not up to the standard, an alarm signal is sent to the numerical control machine, so that the numerical control The machine pauses to process parts.

Specifically, through the custom handshake protocol, the system realizes the real-time reception of measurement data points. After point conversion is calculated into part size, it is compared with the dimensional tolerance specified in the quality standard document in the system, and a report on whether each size meets the standard is output. The information contained in the report includes the measurement time of a part, the code of the part piece, the serial number of the part processing and testing machine, the standard value and tolerance of the part size, the actual value of the part size, the judgment of whether the size meets the standard, and the size deviation value. According to the measurement program, the point measurement value is collected in real time. Real-time judgment whether the measurement size is qualified or not.

Therefore, the above-mentioned system display results not only include the final calculated size result value, but also include the "standard determination" (OK or NG) and the deviation value. By looking at the final collection results, you can immediately know how many dimensions have been measured for the entire piece, whether they all pass, how many dimensions are not up to standard, what is the deviation of each dimension, and which dimensions have deviations exceeding a certain range. If there is a size that does not meet the standard, in addition to displaying it in the system test results, the machine will be notified, causing the CNC machine to stop and generate an alarm to notify the on-site personnel that the processing measurement of this equipment has not passed and needs to be compensated. In order to avoid the next substandard processed product.

S3: Establish a weighted optimization problem model based on the trust region according to the measurement data and measurement model parameters in the standard format, and solve the optimization problem to obtain the optimal compensation value of the CNC machine;

In the case that the size does not meet the standard, the system needs to calculate the compensation value of the tool and the coordinate axis related to the processing of the non-standard size, so that in the next machining process of the machine, after the tool compensation and coordinate system compensation are used, the produced size meets the standard parts. In the process of calculating the compensation value, due to the correlation between the dimensions, at the same time, a tool offset or a change of a coordinate axis may affect multiple dimensions, because the calculation of the compensation value cannot be simply calculated through the inequality of a single dimension. In this embodiment, dimensions, standard values, tolerances, compensation systems, unknown compensation variables, and compensation coefficients of compensation variables are first created to form a set of underdetermined equations, and considering the importance of different sizes to achieve tolerance standards, weighted The solution method finally converts this system of equations into a constrained optimization problem and solves it. With the help of this compensation value algorithm, it can help the machine to obtain the optimal solution, and achieve the best value for each size, not just reach the standard. The specific compensation algorithm is explained in detail as follows.

S31: Build a model.

In CNC machine compensation, the actual measured value should be within the tolerance range of the standard value plus the tool compensation value (compensation value), so the actual measured value should be equal to the standard value plus the tool compensation coefficient multiplied by the tool compensation plus floating tolerance. Among them, the tool compensation coefficient and the standard value are given by the system and obtained in step S1, and the measured value is given by the on-site numerical control machine measurement, and the compensation value and tolerance need to be obtained. Therefore, in the compensation algorithm, the actual physical quantity known to be measured is recorded as C∈R ⁿ , the compensation coefficient z∈R ^k×n , the standard value S∈R ⁿ , the unknown tolerance ε∈R ⁿ and the compensation value X∈R ^k ( Where n and k are both positive integers, representing the dimension of the space, and R represents the real number space, for example, R ⁿ represents the n-dimensional real number space). So there are equations:

C＝S+z ^T X+ε (1)

The unknowns are the compensation value X and the tolerance ε, so the number of unknowns is k+n, and the number of equations is n, so it can be known that the number of unknowns is greater than the number of effective equations, so the problem is transformed into underdetermined System of equations solving problems.

Further, simplify the equation system to the standard form, x=[X,ε]∈R ^n+k , A=[z,I] ^T ∈ R ^n×(n+k) , b=CS∈R ⁿ , Since there are range requirements for the compensation value and tolerance in the actual situation, the present invention equates the original problem to solving the underdetermined equations with constraints:

Among them, lsp, usp∈R ^n+k are the upper bound and lower bound of the unknown quantity respectively.

If there is a feasible solution to the equation group (2), its solution must be the optimal solution of the following constrained optimization problem:

s.t.lsp≤x≤usp

(3) In practice, there are often more requirements for feasible solutions, such as finding the least square norm solution or weighted solution. Therefore, in the present invention, the problem is modeled as a weighted optimization problem solution:

s.t.lsp≤x≤usp

Ax＝b (4)

Here E∈R ^n×(n+k) is the weight.

S32: Model solution: The above quadratic programming problem can be solved by interior point method or trust region reflective least squares method. In the experiment, due to the small value, the solution of interior point method often has errors, so the trust region reflective least squares method is used. method to solve.

S33: Trust Region Reflection Method:

Consider the unconstrained minimization problem, minimizing f(x): R ⁿ → R. Assuming that you are now at point x in n-dimensional space, you need to find a point with a smaller function value. The basic idea is to approximate f with a simpler function q, which needs to fully reflect the behavior of function f in the neighborhood N of point x. This neighborhood N is the trust region N. The trial steps s are computed by minimizing (or approximately minimizing) over the trust region N. The following are trust region subproblems:

min _s {q(s),s∈N} (5)

If f(x+s)<f(x), the current point is updated to x+s; otherwise, the current point remains unchanged, the trust region N shrinks, and the algorithm calculates the trial step again.

In the process of defining a specific trust region method to minimize f(x), the key issues are how to choose and compute the approximation q (defined at the current point x), how to choose and modify the trust region N, and how to exactly solve the trust region sub question. In standard trust region methods, the quadratic approximation q is defined by the first terms of the Taylor approximation of f at x; the neighborhood N is usually spherical or elliptical. Expressed in mathematical language, the trust region subproblem is usually written as

Among them, g is the gradient of f at the current point x, H is the Hessian matrix (symmetric matrix of the second derivative), D is the diagonal scaling matrix, Δ is the radius of the trust region, which is a positive scalar, ||.||is 2-norm. Such second-order algorithms (algorithms involving finding the Hessian matrix and inversion, which are time-consuming, labor-intensive and computationally complex in large-scale solutions) usually involve calculating all the eigenvalues of H, and applying Newton's method to the following characteristic equations

Equation (7) can provide an exact solution to Equation (6); the eigenvalues of H are generally calculated using eigenvalue decomposition, and the optimal iterative format for solving Equation (6) requires calculating the eigenvalue decomposition of H, and applying these eigenvalues and Δ calculates the iterative direction, which involves a theorem for solving quadratic programming problems with constraints:

s ^* is the global minimum solution of formula (6) if and only if s ^* is feasible and there exists λ≥0 such that:

(H+λI)s ^* ＝-g

λ(Δ-||s ^* ||)＝0

(H+λI)≥0

According to the above theorem, it can be known that the iterative format s(λ)=-(H+λI) ^-1 g is a family of vectors about λ, and the optimal s ^* can be found by calculating the optimal λ ^* .

Suppose H has eigenvalue decomposition H=QΛQ ^T , where Q=[q ₁ ,q ₂ ,…,q _n ] is an orthogonal matrix, Λ=Diag(λ ₁ ,λ ₂ ,…,λ _n ) is a diagonal matrix , λ ₁ ≤λ ₂ ≤…≤λ _n are the eigenvalues of H. For convenience, only the case where λ ₁ ≤ 0 and λ ₁ is a single characteristic root is considered below, other cases can be analyzed similarly. Obviously, H+λI has an eigenvalue decomposition H+λI=Q(Λ+λI)Q ^T . For λ>-λ ₁ ≥0, the expression of s(λ) can be written directly:

This is exactly the orthogonal decomposition of s(λ), and it can be easily obtained from the orthogonality

According to the above formula, it can be seen that when λ>-λ ₁ and

When , ||s(λ)|| ² is a strictly decreasing function of λ, and there is

According to the intermediate value theorem of continuous functions, it is known from the complementary relaxation properties that the solution of ||s(λ)||=Δ (equivalent to formula (7)) must exist and be unique, so the unary equation of formula (7) can be calculated by Newton’s method The root of λ ^* is obtained, and then calculated to obtain s ^* .

However, solving the eigenvalues of H takes time proportional to several decompositions of H. Therefore, for the trust region problem, another method is adopted in the present invention. This method is to limit the trust region subproblem within the two-dimensional subspace K for approximate solution. Once the two-dimensional subspace K is calculated, even if a complete For the eigenvalue/eigenvector information, the workload of solving formula (6) is not large, so the main work has been transferred to the determination of the subspace.

The two-dimensional subspace K is determined by means of the following preconditioned conjugate gradient method. The solver defines K as the linear space determined by k ₁ and k ₂ , where k ₁ is the direction of the gradient g and k ₂ is the approximate Newton direction (i.e. the solution of H k ₂ = -g) or the direction of negative curvature

The idea of choosing K in this way is to force global convergence (via the direction of steepest descent or negative curvature) and achieve fast local convergence (via Newton steps, if it exists).

Based on the above analysis, the framework of unconstrained minimization based on trust region can be given:

B1: Construct a two-dimensional trust region subproblem.

B2: Solve equation (6) to determine the trial step s.

B3: If f(x+s)<f(x), then x=x+s.

B4: Adjust the trust region radius Δ.

The trust region radius Δ is adjusted according to standard rules. Specifically, it decreases when a trial step is not accepted (i.e. f(x+s) ≥ f(x)). Repeat the four steps B1 to B4 until the algorithm converges, and then the solution of the optimization problem f(x) (compensation value X and tolerance ε) is obtained.

In the above-mentioned compensation algorithm, it can be seen that the compensation value is determined by the standard value of the size, the tolerance interval, the compensation coefficient, and the size relationship. Therefore, the present invention designs a set of linear equations by modeling, and the compensation value of the unknown quantity and the tolerance As a variable, a set of underdetermined equations is formed by combining the above corresponding parameters. Then, according to the importance of the variables, an optimal weighted algorithm is designed to convert the solution of the underdetermined equations into a constrained optimization problem. The above algorithm helps to solve the best compensation difference, but this compensation difference is not the tool compensation value or coordinate system compensation value that is finally written into the machine. Considering that the compensation may be completed in batches, or considering that the machining itself has already used a certain amount of compensation, it is necessary to consider obtaining the original tool compensation and original coordinate system compensation values used in this collection itself, and then combine it with this time Compensation differences are calculated and summed up, which is the machine compensation value required to produce standard-sized parts. Therefore, at the beginning of the process of handshaking and receiving data, the intelligent machine tuning system has already carried out a collection of corresponding tool compensation and coordinate system compensation according to the standard documents of the currently processed parts. And store the data to prepare for the compensation value to be written into the machine.

S4: Send the optimized compensation value back to the CNC machine, so that the CNC machine can adjust the machine according to the optimized compensation value.

During measurement, the initial value of tool compensation (including radius tool compensation wear and length tool compensation wear) of each tool position and the initial value of each coordinate axis will be collected. Calculate the compensation difference through an algorithm. Finally, the compensation value is written back by adding the compensation difference to the initial value. In addition, during the write-back process of compensation, the system creatively performs fool-proof and error-proof processing, and sets a threshold range. Once the compensation value is found to exceed the threshold, it will not be able to make up for it, and a prompt will be issued to inform that compensation of all sizes that meet the standard is required this time. The value may cause the danger of machine processing, prompting manual intervention to solve it. For example, the absolute value of the compensation value is 0.1 as the limit. Once it is determined that the calculated compensation value combined with the original machine acquisition compensation value exceeds this value, it cannot be directly written back to the machine through system compensation. The security scope can be customized according to the scene requirements. If all compensation values are within the safe range, it can be written back with one key, which is convenient and quick. According to this process, the adjustment of the machine only needs two processing and measurement in total. The first measurement helps to calculate the compensation parameters of the equipment, and the second processing can complete the production of a part with all dimensions up to standard.

The preferred embodiment of the present invention also discloses an intelligent machine adjustment system based on optimization compensation, which is used to perform intelligent machine adjustment on at least one numerical control machine, including: a processor and a storage medium, and the computer program is stored in the storage medium , the processor is configured to run the computer program to execute the above-mentioned intelligent tuning method.

In terms of compensation schemes, at present, in addition to relying on the experience of technicians to adjust the machine, some large-scale enterprises also have some semi-automatic machine tool parameter adjustment schemes. The basic processing flow is: take the workpiece to be measured to the measurement room for measurement by three-coordinate and other measuring instruments , and select a part of the measurement results to manually enter into an excel file, or a system with an algorithm. In Excel documents or systems, experience-based "geometric formulas" will be established, through which the results of parameter adjustment can be calculated relatively quickly and uniformly. This calculation scheme solves the problem of standardization to a certain extent. But there is still a lot of room for improvement in terms of "efficiency" and "accuracy". The machine tuning scheme of the preferred embodiment of the present invention is a set of "automation and intelligence" integration, and a higher degree of scheme: first, the automatic and real-time reading of measurement data is realized by "in-machine measurement"; In this process, the possibility of machine processing defective products during laboratory measurement and waiting time is avoided. Secondly, a complete model is also established for the measurement data and parameters, and the problem of geometric calculation is transformed into an "optimization problem", in which various optimization algorithms are used to solve the problem. This solution has greatly improved the reliability of ensuring the accuracy of all size adjustments.

In the process of machine production and processing, precision testing equipment will be used to conduct product quality testing inside the machine, outside the machine, or both at the same time, and obtain the test results of each process and each size of the product. The intelligent tuning system will obtain various testing data in real time by interacting with testing equipment. And calculate the test data to determine whether each size meets the standard or not. If all dimensions meet the standard, the product has passed the quality verification and can be delivered. If there is a size out of tolerance, the system will calculate the best tool compensation or the best coordinate system compensation for machining of this size through the optimization algorithm. At the same time, the intelligent adjustment system will directly communicate with the processing equipment, and write the best adjustment data into the equipment to complete the adjustment. Compared with the previous manual process of repeatedly debugging errors and repeatedly adjusting compensation schemes based on experience, and finally achieving the quality inspection of all dimensions, the intelligent adjustment system improves the adjustment efficiency by 80%, calculates the optimal result at one time, and ensures that all dimensions pass the inspection. , and through direct interaction between the system and testing equipment, and processing equipment, the testing and processing process is optimized and manual misoperation is prevented. In addition, whether the product quality meets the standard is not only related to the standard process, but also has a lot to do with the performance characteristics of the processing equipment and the experience of the adjustment personnel. After using the intelligent adjustment system for adjustment operation, the quality of equipment and adjustment personnel Dependence will disappear, which once again improves efficiency and reduces the cost of machine adjustment.

The method and system for intelligent machine adjustment based on optimization compensation of the present invention will be further described below in conjunction with specific embodiments.

As shown in Figure 3, it is a flow chart of an intelligent machine tuning method according to a specific embodiment of the present invention, which specifically includes the following steps:

C1: Modus generates a measurement program. This module is the pre-work for the intelligent machine tuning system to start working. Since the machine tuning platform needs to collect measurement content, here, the generation process of the measurement program will be introduced first. Modus is the supporting software system for Renishaw's measuring probes. Its work content is "input the part processing drawing, mark the measurement size, mark the measurement point, and generate the corresponding measurement program". Therefore, the first step can cooperate with software such as Modus to output the NC program of the measurement point. This step itself does not require the intervention of the machine tuning platform, but the process of running the measurement NC program in the machine needs to establish a handshake with the acquisition program of the intelligent machine tuning system.

C2: Import the measurement model and quality parameters into the IMIQ system, and import the basic data into the intelligent adjustment system for subsequent compensation calculations. The imported data content includes: project, part, process, process method, quality standard.

C3: followed by "Receive", which contains "Receive part measurement point" and "Receive sensor or measurement block data". This part does not require manual intervention. The measurement program will be automatically started during the machining process of the machine, and the acquisition program is resident and running on the edge acquisition hardware in the form of a service, and the measurement data is collected through a specific protocol during the automatic measurement process of the machine. And the measurement block data at the corresponding time point are stored and reported.

C4: For all the collected data, in this step, the first step of calculation conversion will be performed, and the point data will be converted into size data. The first is to store all the point data of the entire part, and then interface with Modus for transmission. Calculate the size with the help of the existing point position and size relationship of Modus. Of course, when collecting the original point value, the intelligent adjustment system will consider the point deformation of the standard block to process the original point (by calculating the deformation coefficient of the measurement block to calculate the coordinate offset of each measurement point), and then use calculate.

C5: Optimizing and solving the machine parameters is obtained by the machine tuning platform algorithm.

C6: The last part is the compensation value write-back part. In order to prevent fools and mistakes, the intelligent machine adjustment system itself will have a set of inspection mechanisms to filter out compensation results that are greater than the safety threshold. Subsequently, it is necessary to manually confirm the compensation and write it back to the machine.

As shown in FIG. 4 , it is a frame diagram of an intelligent machine tuning system according to a specific embodiment of the present invention. The system adopts a B/S-based hybrid computing architecture. Users can complete all operations of the system based on the browser. The system computing is completed through the central high-performance server cluster (providing a high-availability mechanism and linear expansion) and cooperating with the hardware and software of the edge computing.

There are currently two access methods for the device:

1) Lightweight access. The CNC machine connects the equipment to the computing network through a network cable or wireless router. The central server is responsible for communicating with all CNC machines. After the measurement data is collected, the parameters are calculated and the results are fed back to the CNC machine.

2) A smarter access solution. It is suitable for scenes with many measurement points and high measurement accuracy requirements. The corresponding Edge (edge controller) will be deployed on the machine side and directly connected to the machine. Edge is responsible for communicating with the measurement program of the CNC machine and implementing the data collection of each group of sensors. All the data is aggregated and simply calculated on the Edge (mainly including matching with the meta data of the part measurement, verification and data completion, etc.) and then sent to the central server for subsequent processing.

As shown in Figure 4, the intelligent machine adjustment system of this specific embodiment includes a controlled CNC machine 10 and a corresponding probe 11, a measurement CAM server (Modus) 20, an access switch 30, a client 40, an edge computing server 50, The specific applications of wireless AP 60 and intelligent terminal 70 are as follows:

1) Manually judge which dimensions can be measured in the CNC machine according to the measurement standard file and the 2D map of the measurement point, generate the in-machine measurement point file and upload it to the system.

2) Use the Modus software to open the 3D map of the corresponding clamping position, mark the corresponding measurement point on the 3D map according to the measurement point file in the machine, and generate the measurement CNC program.

3) The external edge computing server box (including processors, IO modules, reserved sensor interfaces, etc.) is directly connected to the machine tool. And the measurement point data collected on the high-speed receiver station through a specific handshake protocol.

4) The thermal deformation caused by temperature will have a greater impact on the measurement results of the workpiece. In order to further improve the measurement accuracy. The corresponding standard blocks are placed in the CNC machine, and the measurement data of the standard blocks will be collected while the workpiece is measured in the machine. And use the relevant empirical formula to correct the result value of the workpiece measurement, as shown in Figure 5.

5) After the data collection of all workpiece measurement points is completed, it will pass through the DMIS data service (DMIS data service refers to the background service that can support the calculation of dimensions or tolerances defined in the commonly used <Dimension Measurement Interface Standard>, such as Renishaw's modus service , internal DMIS data service) into standard measurement data (supports different standards. At present, the internal standard is used in the communication protocol. Because it is more concise and has better performance. But it can also be converted into industry-wide standards, such as QIF).

6) After the result data of 3), 4) and 5) are summarized and organized, they will be filled into the pre-established processing data model. And use various optimization algorithms to calculate the optimal values of various tuning parameters. The specific algorithm formulas are detailed below.

7) The adjustment parameters may be automatically written back to the CNC machine to affect the next processing. The write-back has also been dealt with fool-proof and error-proof (according to the rules set manually).

8) The model of the off-line test carried out by the three-coordinate measuring instrument in the laboratory can also be imported (manually, automatically) into the system of the present invention and the optimization algorithm of the present invention is also used to give the optimal value of the tuning parameters.

The above content is a further detailed description of the present invention in conjunction with specific preferred embodiments, and it cannot be assumed that the specific implementation of the present invention is limited to these descriptions. For those skilled in the art to which the present invention belongs, several equivalent substitutions or obvious modifications can be made without departing from the concept of the present invention, and those with the same performance or use should be deemed to belong to the protection scope of the present invention.

Claims (15)

1. An intelligent adjustment method based on optimal compensation, which is used for intelligent adjustment of at least one numerical control machine, is characterized in that it includes the following steps:

   S1: Obtain the measurement model parameters of the CNC machine, and receive the actual measurement data of the processed parts collected by the CNC machine;

   S2: Convert the actual measurement data of the processed parts collected by the CNC machine into measurement data in a standard format according to the measurement model parameters;

   S3: Establish a weighted optimization problem model based on the trust region according to the measurement data and measurement model parameters in a standard format, and solve the optimization problem model to obtain the optimal compensation value of the numerical control machine;

   S4: Send the optimized compensation value back to the numerical control machine, so that the numerical control machine performs machine adjustment according to the optimized compensation value.
2. The intelligent adjustment method according to claim 1, characterized in that the measurement model parameters obtained in step S1 include the types, standard values and tolerance standards of the dimensions to be measured of the machined parts of the numerical control machine, the numerical control Tool compensation coefficient of the machine.
3. The intelligent adjustment method according to claim 1, characterized in that step S1 further comprises: correcting the actual measurement data of the processed parts collected by the numerical control machine by using point deformation of the standard block.
4. The intelligent adjustment method according to claim 3, wherein the correction of the actual measurement data collected by the numerical control machine by using the point deformation of the standard block specifically includes: providing the corresponding dimensions of the processed parts to be measured Each standard block is provided with a plurality of calibration points, and the reference value of each calibration point of the standard block corresponding to each size to be measured is obtained; when receiving the actual measurement data of the processed parts collected by the numerical control machine , and also receive the calibration value of each calibration point of the standard block corresponding to each size to be measured, and process the parts collected by the CNC machine according to the reference value and calibration value of each calibration point of the standard block corresponding to each size to be measured The actual measured data is corrected.
5. The intelligent machine tuning method according to claim 1, wherein in step S1, receiving the actual measurement data of the processed parts collected by the numerical control machine is specifically: receiving the data by establishing a handshake agreement with the numerical control machine The actual measurement data of the processed parts collected by the numerical control machine.
6. The intelligent adjustment method according to claim 5, characterized in that, receiving the actual measurement data of the processed parts collected by the numerical control machine by establishing a handshake agreement with the numerical control machine specifically includes:

   A1: Establish a long link with the CNC machine to notify the CNC machine to enter the collection state when receiving data;

   A2: Receive the feature number transmitted by the CNC machine in the form of macro variables and a corresponding point measurement data collected;

   A3: Reset the macro variable, and judge whether the CNC machine has collected all the points, if yes, complete the reception, if not, notify the CNC machine to enter the measurement and reporting of the next point, and return Step A2.
7. The method for intelligent machine adjustment according to claim 1, characterized in that step S2 further includes: judging whether the measured data in a standard format obtained according to the measurement model parameters is up to standard, and if not up to standard, sending an alarm signal to the A numerically controlled machine, such that the numerically controlled machine stops processing parts.
8. According to the intelligent tuning method according to any one of claims 1 to 7, it is characterized in that step S3 specifically includes: the weighted optimization problem model based on the trust region established according to the measurement data in the standard format and the measurement model parameters is expressed as follows:

   

   s.t.lsp≤x≤usp

   Ax=b

   And using the trust region reflection method to solve the above-mentioned optimization problem model to obtain the optimal compensation value of the numerical control machine;

   Among them, A=[z,I] ^T ∈ ^{R n×(n+k)} , x=[X,ε]∈R ^n+k , b=CS∈R ⁿ , C∈R ⁿ is the measurement data in the standard format , S∈R ⁿ is the standard value of each dimension of the processed part included in the measurement model parameters, z∈R ^k×n is the tool compensation coefficient of the CNC machine included in the measurement model parameters, ε∈R ⁿ is the The tolerance of the CNC machine to be solved, X∈R ^k is the compensation value of the CNC machine to be solved, I is the unit matrix, E∈R ^n×(n+k) is the weight, lsp, usp∈R ^n+k are respectively are the upper and lower bounds of the unknown quantity x.
9. The intelligent machine tuning method according to claim 8, characterized in that, using the trust region reflection method to solve the above-mentioned optimization problem model to obtain the optimal compensation value of the numerical control machine specifically includes:

   Construct function f(x), and calculate X and ε by minimizing f(x) on trust region N through trial steps s, where trust region N is the neighborhood of function f(x) at point x.
10. The intelligent tuning method according to claim 9, wherein minimizing f(x) on the trust domain N through the trial step s specifically includes:

    B1: Construct a two-dimensional trust region subproblem:

    

    Among them, g is the gradient of the function f at the current point x, H is the Hessian matrix, D is the diagonal scaling matrix, s is the trial step, and Δ is the radius of the trust region;

    B2: Solve the two-dimensional trust region subproblem to determine the trial step s;

    B3: Determine whether f(x+s) is smaller than f(x), if yes, make x=x+s, and return to step B1; if not, keep the current point unchanged, reduce the trust region radius Δ, and return Step B1;

    B4: Repeat steps B1-B3 until the two-dimensional trust region subproblem converges, and the solution to the optimization problem f(x) is obtained.
11. The intelligent tuning method according to claim 10, characterized in that, in step B2, when solving the two-dimensional trust region subproblem, the two-dimensional trust region subproblem is limited to the two-dimensional subspace K for solving, and the two-dimensional subspace The space K is determined according to the following preconditioned conjugate gradient method: the two-dimensional subspace K is defined as a linear space determined by k ₁ and k ₂ , where k ₁ is the direction of the gradient g, and k ₂ satisfies H·k ₂ =- g or

    

    conditions of.
12. The intelligent tuning method according to claim 8, characterized in that, after solving the above-mentioned optimization problem model to obtain the optimal compensation value of the numerical control machine by adopting the trust region reflection method, it also includes: obtaining according to the solution optimization problem model The optimized compensation value of the numerical control machine and the measurement model parameters correct the optimal compensation value of the numerical control machine.
13. The intelligent machine tuning method according to claim 1, characterized in that before sending the optimized compensation value back to the numerical control machine in step S4, it is judged whether the optimized compensation value exceeds a preset threshold, and if so, Then send an intervention signal to the numerical control machine, so that the numerical control machine stops processing parts; if not, then send the optimized compensation value back to the numerical control machine.
14. The intelligent machine tuning method according to claim 1, characterized in that before sending the optimized compensation value back to the numerical control machine in step S4, it is judged whether the optimized compensation value exceeds a preset threshold, and if so, Then use the standard block again to correct the actual measurement data of the processed parts collected by the CNC machine; if the correction still exceeds the threshold, send an intervention signal to the CNC machine so that the CNC machine suspends processing the parts ; If the preset threshold value is not exceeded, the optimized compensation value is sent back to the numerical control machine.
15. An intelligent machine tuning system based on optimal compensation, used for intelligent machine tuning of at least one numerical control machine, is characterized in that it includes: a processor and a storage medium, and a computer program is stored in the storage medium, and the processing The device is configured to run the computer program to execute the intelligent tuning method according to any one of claims 1 to 14.

Images