Search

CN-121635070-B - Numerical control device and method for process in-loop

CN121635070BCN 121635070 BCN121635070 BCN 121635070BCN-121635070-B

Abstract

The invention discloses a numerical control device and a numerical control method for a process in a loop, which belong to the technical field of numerical control and comprise a non-real-time process optimizer, wherein the non-real-time process optimizer performs process optimization processing on original track data input by a user based on a track error prediction model of deep learning, effectively compensates for comprehensive errors generated in a multi-axis machining process, an interpolator and a position controller are used for calculating coordinates of a middle point of a motion track and accurately controlling a tool to reach a target position, the process prediction observer utilizes a neural network technology to construct a dynamic observer for real-time feedforward compensation in the machining process, and a physical machine tool performs machining operation. The invention adopts the numerical control device and the numerical control method of the process in the loop, adds a non-real-time process optimizer and a process prediction observer, predicts, optimizes and dynamically controls the processing process in a closed loop by means of the depth fusion of virtual machine simulation and physical machine feedback, optimizes the processing track in real time and compensates errors.

Inventors

  • SHEN BIN
  • HUANG YUNYING
  • CHEN SULIN

Assignees

  • 上海交通大学

Dates

Publication Date
20260505
Application Date
20260204

Claims (4)

  1. 1. The numerical control device of the process in the ring is characterized by comprising a non-real-time process optimizer, an interpolator, a position controller, a process prediction observer and a physical machine tool; the non-real-time process optimizer performs process optimization processing on original track data input by a user based on a track error prediction model of deep learning, and effectively compensates for comprehensive errors generated in the multi-axis machining process; The interpolator and the position controller are used for calculating the coordinates of the middle point of the motion trail and accurately controlling the cutter to reach the target position; the process prediction observer utilizes a neural network technology to construct a dynamic observer, and performs real-time feedforward compensation in the processing process; The physical machine tool performs machining operation; The numerical control method of the numerical control device for the process ring comprises the following steps: s1, preprocessing programming track data by a non-real-time process optimizer to obtain a track data stream; s2, inputting the track data stream into an interpolator and a position controller, and controlling a physical machine tool to move the position; s3, inputting an actual moving track of the physical machine tool as a target track into a process prediction observer of the physical machine tool and a control model of the virtual machine tool to obtain an observation error, and correcting and obtaining a fine-tuned periodic track by a track error compensator; the acquisition of the fine-tuned periodic track specifically comprises the following steps: s3.1, inputting a theoretical periodic track into an NN neural network to acquire a control signal; Using a multi-layer neural network model Servo motor system of fitting machine tool and input system state vector Outputting control instruction The following is shown: ; Wherein, the Is the first The position error of the step is determined, In order to be a speed error, In order for the current error to be a function of, As a rate of change of the position error, As a rate of change of the velocity error, Is the first Cutting force applied to the step system; the training objective is to minimize the error between the predicted output and the ideal response, as follows: ; Wherein, the In order to cope with the loss function, Is the first In the step of the method, the device comprises the steps of, Is the first The ideal response of the step is that, At parameters for neural network Lower, the first The predicted output of the step(s), In order for the parameters to be regularized, For neural networks with respect to parameters Is a gradient of (2); s3.2, inputting a control signal into the cutting force model and the displacement error model to estimate the disturbance force; S3.3, optimizing and obtaining a fine-tuned periodic track by a track correction optimizer after the disturbance force is evaluated; s4, inputting the acquired fine adjustment periodic track into a physical machine tool, and controlling the operation of the physical machine tool.
  2. 2. The numerical control device of the process in-loop of claim 1, wherein the target track in S3 is input into a process prediction observer of a physical machine tool and a control model of a virtual machine tool to obtain dynamic observation errors, and the dynamic observation errors are estimated deviations of cutting force disturbance and position errors, and the cutting force and the position errors are specifically shown as follows: ; ; Wherein, the In order for the cutting force to be high, For the cutting force correction factor to be applied, In order to achieve a depth of cut, In order to achieve a cutting speed, the cutting speed, For the radius of the tool, For the spindle speed, In order to integrate the position errors, As a coefficient of speed (f) the speed, Is used as a mass coefficient of the composite material, Is a coefficient of proportionality and is used for the control of the power supply, As an integral coefficient of the power supply, Is a differential coefficient.
  3. 3. The numerical control device for the process ring according to claim 2, wherein the calculation formula of the cutting force model in S3.2 is as follows: ; the calculation formula of the displacement error model is as follows: ; Wherein, the For the tool deformation displacement error caused by the cutting force, As the error deformation coefficient, the deformation coefficient, Is the first The step system is subjected to cutting forces.
  4. 4. A numerical control device of a process in-loop according to claim 3, wherein S3.3 evaluates a predicted trajectory error according to the disturbance force, and the predicted formula of the trajectory error is as follows: ; Wherein, the Is the first The predicted trajectory error of the cycle is calculated, Is the first The current trajectory error of the cycle is calculated, Follow-up errors predicted for the neural network; Defining trajectory input modifier The optimization objective is to minimize the prediction error, and the calculation formula is as follows: ; Wherein, the For the correlation matrix of correction amount and error, Is the first The target error for the cycle is determined, Is a regularization parameter; solving the formula to obtain correction quantity compensation as The fine-tuned output is , wherein, Is the original control output.

Description

Numerical control device and method for process in-loop Technical Field The invention relates to the technical field of numerical control, in particular to a numerical control device and method for a process in-loop. Background The existing numerical control device or system generally cannot solve the problem of process closed loop because the process information cannot be closed loop with the numerical control device, the process optimization depends on manual experiments and adjustment, and the process optimization result cannot be reflected in the numerical control device in real time. The traditional cross coupling control mode is limited by the performance of the controller when compensating the multi-axis linkage processing track error, the error adjustment range is limited by the characteristics of the controller, fine adjustment can be performed only in a limited range, and the traditional numerical control system is dependent on an empirical formula and off-line compensation, so that the traditional numerical control system is difficult to cope with dynamic factors such as thermal errors, cutting force disturbance, cutter abrasion and the like in the complex multi-axis processing process. Disclosure of Invention The invention aims to provide a numerical control device and a numerical control method for a process loop, wherein a non-real-time process optimizer and a process prediction observer are added, and effective compensation of comprehensive errors generated in a multi-axis machining process and real-time feedforward compensation in the machining process are realized through deep fusion of virtual machine tool simulation and physical machine tool feedback, so that a machining process is predicted, optimized and dynamically controlled in a closed loop, a machining track is optimized in real time, and the errors are compensated. In order to achieve the above purpose, the invention provides a numerical control device and method for a process in-loop, comprising a non-real-time process optimizer, an interpolator and a position controller, a process prediction observer and a physical machine tool; the non-real-time process optimizer performs process optimization processing on original track data input by a user based on a track error prediction model of deep learning, and effectively compensates for comprehensive errors generated in the multi-axis machining process; The interpolator and the position controller are used for calculating the coordinates of the middle point of the motion trail and accurately controlling the cutter to reach the target position; the process prediction observer utilizes a neural network technology to construct a dynamic observer, and performs real-time feedforward compensation in the processing process; and (5) performing machining operation by a physical machine tool. The invention provides a numerical control method of a process in-loop, which is applied to the numerical control device of the process in-loop and comprises the following steps: s1, preprocessing programming track data by a non-real-time process optimizer to obtain a track data stream; s2, inputting the track data stream into an interpolator and a position controller, and controlling a physical machine tool to move the position; s3, inputting an actual moving track of the physical machine tool as a target track into a process prediction observer of the physical machine tool and a control model of the virtual machine tool to obtain an observation error, and correcting and obtaining a fine-tuned periodic track by a track error compensator; s4, inputting the acquired fine adjustment periodic track into a physical machine tool, and controlling the operation of the physical machine tool. Preferably, in S3, the target track is input into a process prediction observer of the physical machine tool and a control model of the virtual machine tool to obtain a dynamic observation error, where the dynamic observation error is an estimated deviation of cutting force disturbance and position error, and the cutting force and position error are specifically shown as follows: ; ; Wherein, the In order for the cutting force to be high,For the cutting force correction factor to be applied,In order to achieve a depth of cut,In order to achieve a cutting speed, the cutting speed,For the radius of the tool,For the spindle speed,In order to integrate the position errors,As a coefficient of speed (f) the speed,Is used as a mass coefficient of the composite material,Is a coefficient of proportionality and is used for the control of the power supply,As an integral coefficient of the power supply,Is a differential coefficient. Preferably, the acquiring of the fine-tuned periodic track in S3 specifically includes the following steps: s3.1, inputting a theoretical periodic track into an NN neural network to acquire a control signal; s3.2, inputting a control signal into the cutting force model and the displacement error model to estim