Intelligent compensation method for thermal error of numerical control system based on multi-source temperature sensing

Abstract

The present application relates to the technical field of numerical control system control, in particular to a kind of intelligent compensation method for thermal error of numerical control system based on multi-source temperature perception.The method trains two prediction models according to the collected data respectively.The predicted instantaneous temperature residual output by loss temperature residual prediction model can ensure reference while not causing control lag problem;Machine tool background temperature prediction model is a low-frequency model with one sampling period as time unit, and planned energy injection and open-loop thermal instability degree are taken as driving features of machine tool heat source physical properties in sampling period to predict predicted machine tool background temperature in future period.According to the prediction results of the two models, the laser power and machine tool feed rate can be compensated and controlled.The present application avoids the technical problems of transient overheating control lag and not considering strong heat source in load increasing process by effective training and application of two prediction models.

Description

Technical Field

[0001] This invention relates to the field of CNC system control technology, specifically to an intelligent thermal error compensation method for CNC systems based on multi-source temperature sensing. Background Technology

[0002] In metal additive manufacturing, high-energy-density heat sources induce two types of thermal errors on different time scales within the machine tool system, jointly constraining the final machining accuracy. First, on a fast time scale of milliseconds to seconds, when the heat source scans areas of the part with limited heat dissipation, such as sharp corners or thin walls, localized heat accumulation occurs rapidly, potentially leading to transient thermal deformation such as overheating and melting of the workpiece. Second, on a slow time scale of minutes to hours, the heat dissipated throughout the additive manufacturing process continuously accumulates and is conducted to critical structural components such as the machine tool bed and column, causing uneven thermal expansion and resulting in macroscopic positioning drift at the tool center point. These two error sources are distinctly different in their causes and time scales, yet they are interconnected.

[0003] Existing technologies have limitations in addressing such problems, focusing on error compensation at a single scale and struggling to effectively handle complex conditions with coexisting dual-scale errors. Control of transient overheating in workpieces often relies on delayed temperature feedback or fixed process parameters, making it difficult to prevent overheating caused by complex geometry and machining paths. Regarding compensation for macroscopic thermal drift in machine tools, current methods generally consider the machine tool's own temperature field in isolation, neglecting the direct impact of the strong heat source in the additive manufacturing process as the actual driving force on the machine tool's thermal deformation. Summary of the Invention

[0004] To address the shortcomings of existing technologies in transient overheat control and the lack of consideration for strong heat sources during machine tool thermal drift compensation, this invention aims to provide an intelligent thermal error compensation method for CNC systems based on multi-source temperature sensing. The specific technical solution adopted is as follows:

[0005] This invention proposes an intelligent thermal error compensation method for CNC systems based on multi-source temperature sensing, the method comprising:

[0006] The static heat dissipation characteristics of the CNC machine tool machining task are obtained based on the task information; during the execution of the machining task, the command power and instantaneous molten pool temperature are collected at each moment in the preset sampling period; temperature readings at different locations on the characteristic structural parts of the CNC machine tool are collected to obtain the machine tool structure temperature feature vector and the machine tool background temperature.

[0007] At each time point, the processing temperature is obtained based on the machine tool background temperature and command power, and the instantaneous temperature residual at each time point is obtained based on the difference between the instantaneous molten pool temperature and the processing temperature. The machine tool structure temperature feature vector, command power and static heat dissipation characteristics are used as input data, and the instantaneous temperature residual at the next time point is used as output data to train the instantaneous temperature residual prediction model.

[0008] The planned injected energy in each sampling period is calculated based on the command power, and the open-loop thermal instability degree is obtained based on the instantaneous temperature residual change characteristics in each sampling period. The command power, planned injected energy and open-loop thermal instability degree in the sampling period are used as input data, and the machine tool background temperature in the next sampling period is used as output data to train the machine tool background temperature prediction model.

[0009] The instantaneous temperature residual prediction model is used to process the data under the real-time sampling period to obtain the predicted instantaneous temperature residual; the predicted machine tool background temperature residual is obtained based on the difference between the machine tool background temperature under the real-time sampling period and the predicted machine tool background temperature output by the machine tool background temperature prediction model.

[0010] The laser power is adjusted based on the predicted instantaneous temperature residual, and the machine tool feed rate is adjusted based on the predicted instantaneous temperature residual and the predicted machine tool background temperature residual.

[0011] Furthermore, the static heat dissipation characteristics include geometric heat dissipation conditions and path reheating risk; the volume ratio of the solidified entity to the powder to be melted in the local area of ​​each path point is obtained, and the average of all volume ratios is taken as the geometric heat dissipation conditions; the number of reheating times at each path point is counted, and the average of the number of reheating times at all path points is taken as the path reheating risk.

[0012] Furthermore, the machine tool background temperature is the average value of all elements in the machine tool structure temperature feature vector.

[0013] Furthermore, the method for obtaining the processing temperature includes:

[0014] The input temperature is obtained by multiplying the command power by a preset conversion coefficient, and the sum of the input temperature and the machine tool background temperature is taken as the processing temperature.

[0015] Furthermore, the method for obtaining the instantaneous temperature residual includes:

[0016] The difference between the instantaneous molten pool temperature and the processing temperature is taken as the instantaneous temperature residual.

[0017] Furthermore, the degree of open-loop thermal instability is the standard deviation of the instantaneous temperature residuals at all time points within the sampling period.

[0018] Furthermore, the method for obtaining the predicted machine tool background temperature includes:

[0019] The instantaneous temperature residual prediction model is used to obtain the predicted instantaneous temperature residual sequence for the real-time sampling period. The predicted open-loop thermal instability degree is obtained based on the predicted instantaneous temperature residual sequence. The predicted open-loop thermal instability degree, the planned injection energy under the real-time sampling period, and the command power under the real-time sampling period are used as input data, and the predicted machine tool background temperature is output through the machine tool background temperature prediction model.

[0020] Furthermore, the predicted machine tool background temperature residual is the difference between the predicted machine tool background temperature and the machine tool background temperature under the real-time sampling period.

[0021] Furthermore, adjusting the laser power based on the predicted instantaneous temperature residual includes:

[0022] Divide the predicted instantaneous temperature residual by a preset conversion coefficient to obtain the compensation power; subtract the compensation power from the laser's base power to obtain the adjusted laser power.

[0023] Further, adjusting the machine tool feed rate based on the predicted instantaneous temperature residual and the predicted machine tool background temperature residual includes:

[0024] The sum of the predicted instantaneous temperature residual and the predicted machine tool background temperature residual is used as the cumulative value of the machine tool thermal error factor. If the cumulative value of the machine tool thermal error factor is greater than a preset first threshold and less than a preset second threshold, the feed rate is reduced by a linear decreasing method. If the cumulative value of the machine tool thermal error factor is greater than or equal to the second threshold, the feed rate is adjusted to a preset minimum value.

[0025] The present invention has the following beneficial effects:

[0026] This invention collects static heat dissipation characteristics before processing and multiple thermal characteristics during processing. Based on the collected data, two prediction models are trained: an instantaneous temperature residual prediction model and a machine tool background temperature prediction model. The loss temperature residual prediction model is a high-frequency model trained on high-frequency data. Therefore, its output predicted instantaneous temperature residual ensures reference accuracy without control lag. Furthermore, during training, the model learns the nonlinear differences in overheating response caused by the same geometric path under different global thermal backgrounds, thus obtaining prediction data that accounts for macroscopic effects. The machine tool background temperature prediction model is a low-frequency model with a sampling period as the time unit. It uses the planned injected energy and the degree of open-loop thermal instability as driving characteristics of the machine tool's heat source physical properties under the sampling period. Therefore, the constitutive relationship between the additive process and machine tool thermal deformation can be learned in the machine tool background temperature prediction model, enabling the model to accurately predict the machine tool background temperature in future periods. Based on the prediction results of the two models, timely and reasonable control of laser power and machine tool feed rate can be achieved. Attached Figure Description

[0027] To more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0028] Figure 1 The flowchart illustrates a method for intelligent thermal error compensation in a numerical control system based on multi-source temperature sensing, as provided in one embodiment of the present invention. Detailed Implementation

[0029] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of a multi-source temperature sensing-based intelligent thermal error compensation method for CNC systems proposed according to the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0030] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0031] The following description, in conjunction with the accompanying drawings, details the specific scheme of the intelligent thermal error compensation method for CNC systems based on multi-source temperature sensing provided by this invention.

[0032] Please see Figure 1 The diagram illustrates a flowchart of an intelligent thermal error compensation method for a numerical control system based on multi-source temperature sensing, according to an embodiment of the present invention. The method includes:

[0033] Step S1: Obtain the static heat dissipation characteristics of the CNC machine tool machining task based on the task information; during the execution of the machining task, collect the command power and instantaneous molten pool temperature at each moment in the preset sampling period; collect temperature readings at different locations on the characteristic structural parts of the CNC machine tool to obtain the machine tool structure temperature feature vector and obtain the machine tool background temperature.

[0034] This invention aims to solve the control lag problem by training an effective prediction model. Therefore, it requires collecting various information about CNC machine tool machining tasks, including historical and real-time information. Historical information is used to train the model, while real-time information is input into the trained model for prediction output. This invention sets the sampling period to 1 second, with 100 time points within one sampling period. Furthermore, for each machining task, this invention acquires both static and dynamic machining information.

[0035] Static features are static heat dissipation features obtained from task information. They are usually characterized by the geometric features of the workpiece or the heat dissipation features corresponding to the machining path in the task design file. They are part of the offline preprocessing stage before the machining task begins. Static heat dissipation features can be used to characterize the basic heat information of the machining task.

[0036] Dynamic features are characteristics acquired in real time during the execution of a machining task. Specifically, they include the command power and instantaneous melt pool temperature at each moment in the sampling period, as well as a machine tool structure temperature feature vector composed of temperature readings at different locations on the machine tool's characteristic structural components during the sampling period. The machine tool background temperature can be estimated based on this feature vector. By acquiring dynamic features, it is possible to determine multiple heat-related information during the execution of a machining task, including the heat injected into the CNC machine tool, the real-time feedback heat, and the machine tool background temperature. Analysis of this information allows for the determination of the heat generation and dissipation methods employed by the machine tool during task execution.

[0037] The machine tool feature structural components can be set as key structural components such as the machine tool bed and column, and specific details will not be elaborated or limited further. The machine tool structure temperature feature vector is the vector composed of the temperatures collected at various locations. It should be noted that the settings of the machine tool feature structural components should be consistent across all machining tasks, that is, the length of the temperature feature vectors of different machine tool structures should be consistent, and the machine tool feature structural components corresponding to elements at the same location should be the same.

[0038] Static and dynamic features provide structured data that preserves macroscopic states, dynamic processes, and static attributes, which can be used to train and analyze predictive models. In this embodiment of the invention, for each machining task, with a sampling period of 1 second, a sequence of command power and an instantaneous melt pool temperature consisting of 100 time points can be obtained; a machine tool structure temperature feature vector and a machine tool background temperature can be obtained; and a static heat dissipation feature of this machining task can be obtained.

[0039] Preferably, in this embodiment of the invention, the static heat dissipation characteristics include geometric heat dissipation conditions and path reheating risk; the volume ratio of the solidified entity to the powder to be melted in a local region of each path point is obtained, and the average of all volume ratios is taken as the geometric heat dissipation conditions; the number of reheating cycles for each path point is counted, and the average of the number of reheating cycles for all path points is taken as the path reheating risk. The geometric heat dissipation conditions can characterize the inherent heat dissipation capacity of a path point determined by its local shape (e.g., sharp corners, thin walls) through volume ratios; the reheating risk characterizes the risk of repeated heat injection at the path point location during task execution. That is, the static heat dissipation characteristics in this embodiment of the invention are a two-dimensional vector.

[0040] It should be noted that the size of the local area of ​​the path point can be determined according to the size of the component corresponding to the specific processing task, without further limitation or elaboration.

[0041] Preferably, in this embodiment of the invention, the machine tool background temperature is the average value of all elements in the machine tool structure temperature feature vector.

[0042] Step S2: At each time point, obtain the processing temperature based on the machine tool background temperature and command power, and obtain the instantaneous temperature residual at each time point based on the difference between the instantaneous molten pool temperature and the processing temperature; use the machine tool structure temperature feature vector, command power and static heat dissipation characteristics as input data, and the instantaneous temperature residual at the next time point as output data to train the instantaneous temperature residual prediction model.

[0043] After collecting information on historical processing tasks using step S1, further extraction of the information is needed to facilitate model training.

[0044] First, a high-frequency instantaneous temperature residual prediction model is trained. This model performs prediction analysis using each moment in the sampling period as the time unit, outputting the predicted instantaneous temperature residual for future moments. The purpose of the output is to obtain the instantaneous temperature residual for future moments to adjust the laser power on the CNC machine tool, thereby ensuring temperature stability. Therefore, during model training, the instantaneous temperature residual for each moment needs to be determined and used as a label for the previous moment during training. In this embodiment of the invention, for each moment of a completed machining task, the machining temperature is first obtained based on the machine tool background temperature and the command power. The machine tool background temperature serves as the temperature basis, and the command power is the power parameter of the machine tool when injecting heat into the workpiece material. Therefore, the theoretical machining temperature can be obtained based on these two data points. Then, the instantaneous temperature residual for each moment is obtained based on the difference between the instantaneous molten pool temperature and the machining temperature. In other words, the instantaneous temperature residual represents the additional temperature increase beyond basic physical laws caused by factors such as static heat dissipation characteristics at the current moment. It should be noted that the additional temperature increase here can also be negative, that is, the temperature decreases. The specific value of the instantaneous temperature residual can represent the comparison between heat generation and heat dissipation in the actual processing process. Therefore, this feature can be used as a basis for laser power adjustment in CNC machine tools.

[0045] For the instantaneous temperature residual prediction model, this embodiment of the invention uses the machine tool structure temperature feature vector, command power and static heat dissipation characteristics as input data, and the instantaneous temperature residual at the next time point as output data for training.

[0046] In this embodiment of the invention, the instantaneous temperature residual prediction model employs a Long Short-Term Memory (LSTM) network model. To guide microscopic predictions from macroscopic states, before processing each new data sequence, the model uses an independently seekable affine transformation layer to generate the initial hidden state and unit state of the machine tool structure temperature feature vector corresponding to that data segment. Subsequently, a sequence containing historical instantaneous temperature residuals, historical command power, and the static heat dissipation characteristics of the corresponding machining task is fed into the network model for training. The loss function used in training is the mean squared error function, meaning the training objective is to minimize the mean squared error between the output predicted instantaneous temperature residual sequence and the corresponding historical instantaneous temperature residual sequence. The specific training method for the LSM network is a well-known technique and will not be elaborated or limited here.

[0047] Preferably, in this embodiment of the invention, the method for obtaining the processing temperature includes:

[0048] The input temperature is obtained by multiplying the command power by a preset conversion coefficient. The sum of the input temperature and the machine tool background temperature is taken as the processing temperature. It should be noted that the preset conversion coefficient can be determined in advance by performing linear regression analysis on the processing data of a simple large-area set of data, or by referring to the laser's instruction manual or other prior knowledge. Specific details are not elaborated or limited here.

[0049] Preferably, in this embodiment of the invention, the difference between the instantaneous molten pool temperature and the processing temperature is used as the instantaneous temperature residual. That is, the instantaneous temperature residual has positive and negative values; a positive value represents that additional heat was generated during the actual process, and a negative value represents that additional heat was generated during the actual process.

[0050] Step S3: Calculate the planned injected energy in each sampling period based on the command power, and obtain the open-loop thermal instability degree based on the instantaneous temperature residual change characteristics in each sampling period; use the command power, planned injected energy and open-loop thermal instability degree in the sampling period as input data, and the machine tool background temperature in the next sampling period as output data to train the machine tool background temperature prediction model.

[0051] The machine tool background temperature prediction model is used to predict the slowly accumulating thermal deformation of machine tools driven by the additive manufacturing process. This thermal deformation is mainly obtained by predicting the machine tool background temperature in future sampling periods and then comparing it with the actual machine tool background temperature. Because existing technologies for predicting machine tool background temperature usually only consider the change in machine tool background temperature itself and do not take into account the fundamental heat source driving the additive manufacturing process, this embodiment of the invention extracts the driving characteristics under the additive manufacturing process. In order to quantify the thermal load of the additive manufacturing process on the machine tool, data from each sampling period are statistically analyzed to obtain two features: planned injected energy and open-loop thermal instability degree, as heat source driving characteristics.

[0052] The planned energy injection can be statistically analyzed based on the command power during the sampling period, i.e., by integrating the command power over time to obtain the energy corresponding to the power signal. The planned energy injection quantifies the total heat planned to be injected into the system during the additive manufacturing process within the sampling period. It should be noted that the numerical integration method for calculating energy is a well-known technique in the art, and its specific details will not be elaborated upon here.

[0053] The degree of open-loop thermal instability is obtained by statistically analyzing the change characteristics of all instantaneous temperature residuals at all points in time within the sampling period. The larger the change characteristics, the greater the drastic fluctuation of the molten pool temperature within the sampling period, which can reflect the impact of the heat input in the current sampling period.

[0054] Preferably, in this embodiment of the invention, the degree of open-loop thermal instability is the standard deviation of the instantaneous temperature residuals at all time points within the sampling period.

[0055] In this embodiment of the invention, the input to the machine tool background temperature prediction model is the command power sequence, planned injection energy, and open-loop thermal instability degree within a sampling period, and the output is the machine tool background temperature for the next sampling period. Similar to the instantaneous temperature residual prediction model, the machine tool background temperature prediction model in this embodiment of the invention also adopts a long short-term memory model, and the loss function is also the mean square error function. The training objective is to minimize the mean square error between the machine tool background temperature of the future sampling period output by the model and the machine tool background temperature corresponding to the label.

[0056] Step S4: Process the data under the real-time sampling period using the instantaneous temperature residual prediction model to obtain the predicted instantaneous temperature residual; obtain the predicted machine tool background temperature residual based on the difference between the machine tool background temperature under the real-time sampling period and the predicted machine tool background temperature output by the machine tool background temperature prediction model.

[0057] The above steps complete the training of two prediction models. In practical use, the two trained models can be loaded, and the data collected in the real-time sampling period can be integrated as input data and input into the two models respectively to obtain the predicted instantaneous temperature residual and the predicted machine tool background temperature. The predicted machine tool background temperature needs to be further compared with the machine tool background temperature in the real-time sampling period to obtain the predicted machine tool background temperature residual, which is used to represent the temperature change of the machine tool background temperature under the influence of thermal drive.

[0058] Preferably, in this embodiment of the invention, because the instantaneous temperature residual prediction model is a high-frequency model compared to the machine tool background temperature prediction model, and the output data of the instantaneous temperature residual prediction model is also high-frequency data, the predicted instantaneous temperature residual, being a high-frequency data, will cause the laser power to undergo a certain compensation fine-tuning. To avoid input data distribution mismatch caused by the compensation intervention of the high-frequency model, which would result in the detected instantaneous temperature residual approaching 0 and losing its physical meaning, this embodiment of the invention performs an online input reconstruction. During the real-time sampling period, the planned injected energy and command power are detected normally, while the open-loop thermal instability degree is obtained through the instantaneous temperature residual prediction model. That is, the instantaneous temperature residual prediction model is used to obtain the predicted instantaneous temperature residual sequence during the real-time sampling period, and the predicted open-loop thermal instability degree is obtained based on the predicted instantaneous temperature residual sequence. The predicted open-loop thermal instability degree, the planned injected energy during the real-time sampling period, and the command power during the real-time sampling period are used as input data, and the predicted machine tool background temperature is output through the machine tool background temperature prediction model. The predicted open-loop thermal instability obtained by this method can simulate the change characteristics of the predicted instantaneous temperature residual sequence that will occur without compensation, thus avoiding data distribution mismatch caused by high-frequency compensation intervention.

[0059] Preferably, in this embodiment of the invention, the predicted machine tool background temperature residual is the difference between the predicted machine tool background temperature and the machine tool background temperature under the real-time sampling period.

[0060] Step S5: Adjust the laser power based on the predicted instantaneous temperature residual, and adjust the machine tool feed rate based on the predicted instantaneous temperature residual and the predicted machine tool background temperature residual.

[0061] Based on the analysis of real-time data in step S4, the predicted instantaneous temperature residual at future time points and the predicted machine tool background temperature residual at future sampling times can be obtained. Compensation control of the CNC machine tool can then be performed based on these two residuals, achieving comprehensive suppression of dual-scale thermal errors.

[0062] The predicted instantaneous temperature residual is a high-frequency feature, corresponding to each time point within the sampling period. This means the laser power adjustment frequency is relatively high, and the system actively suppresses localized overheating of the workpiece caused by set and path factors by adjusting the laser power in real time. In this embodiment, the compensation channel employs a model-based feedforward control. At each 10-millisecond high-frequency moment, the system first calculates a base power command based on the preset target molten pool temperature and the background temperature determined by the current machine tool thermal state. This power is the power required to maintain the target temperature under ideal conditions without any overheating risk. The adjusted laser power can then be obtained by adjusting this base power based on the predicted temperature residual.

[0063] Preferably, in this embodiment of the invention, adjusting the laser power based on the predicted instantaneous temperature residual includes:

[0064] The compensated power is obtained by dividing the predicted instantaneous temperature residual by a preset conversion coefficient; the adjusted laser power is obtained by subtracting the compensated power from the laser's base power. The method for obtaining the conversion coefficient has already been explained in the section on obtaining the processing temperature and will not be repeated here. This adjustment method automatically reduces the power before the laser spot reaches the risk area, based on an overheat prediction that takes into account the global thermal state, thereby effectively preventing the melting and collapse of the part's delicate structure.

[0065] In this embodiment of the invention, the second compensation method uses a sampling period as the time unit, that is, it adjusts the machine tool feed rate at a low frequency to avoid further amplification of mechanical errors under heat accumulation.

[0066] Preferably, in this embodiment of the invention, adjusting the machine tool feed rate based on the predicted instantaneous temperature residual and the predicted machine tool background temperature residual includes:

[0067] The sum of the predicted instantaneous temperature residual and the predicted machine tool background temperature residual is used as the accumulated value of the machine tool thermal error factor. If the accumulated value of the machine tool thermal error factor is greater than a preset first threshold and less than a preset second threshold, it indicates that the feed rate needs to be adjusted, and the feed rate is reduced using a linear decreasing method. If the accumulated value of the machine tool thermal error factor is greater than or equal to the second threshold, it indicates that the thermal error risk is relatively high, and in order to avoid deterioration, the feed rate is adjusted to a preset minimum value. In this embodiment of the invention, the first threshold is set to 75% of the median value in the dimension of the accumulated value of the machine tool thermal error factor; the second threshold is set to the higher value of the top 90% in the dimension of the accumulated value of the machine tool thermal error factor.

[0068] It should be noted that linear descent involves setting a small descent amount, and each adjustment is made using this descent amount as the step size. The specific descent amount can be set according to the specific task, which will not be elaborated upon in this embodiment of the invention.

[0069] In summary, this invention collects static heat dissipation characteristics before processing and multiple thermal characteristics during processing. Based on the collected data, two prediction models are trained: an instantaneous temperature residual prediction model and a machine tool background temperature prediction model. The instantaneous temperature residual output by the loss temperature residual prediction model ensures reference accuracy without causing control lag. The machine tool background temperature prediction model is a low-frequency model with a sampling period as the time unit. It uses the planned injected energy and the degree of open-loop thermal instability as driving characteristics of the physical properties of the machine tool heat source during the sampling period, enabling the model to accurately predict the predicted machine tool background temperature in future periods. Furthermore, based on the prediction results of the two models, the laser power and machine tool feed rate can be controlled in a timely and reasonable manner. This invention, through the effective training and application of the two prediction models, avoids the technical problems of relatively lagging transient overheat control and the failure to consider strong heat sources during the load increase process in machine tool thermal drift compensation.

[0070] It should be noted that the order of the above embodiments of the present invention is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. The processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0071] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

Claims

1. A method for intelligent thermal error compensation in a numerical control system based on multi-source temperature sensing, characterized in that, The method includes: The static heat dissipation characteristics of the CNC machine tool machining task are obtained based on the task information; during the execution of the machining task, the command power and instantaneous molten pool temperature are collected at each moment in the preset sampling period; temperature readings at different locations on the characteristic structural parts of the CNC machine tool are collected to obtain the machine tool structure temperature feature vector and the machine tool background temperature. At each time point, the processing temperature is obtained based on the machine tool background temperature and command power, and the instantaneous temperature residual at each time point is obtained based on the difference between the instantaneous molten pool temperature and the processing temperature. The machine tool structure temperature feature vector, command power and static heat dissipation characteristics are used as input data, and the instantaneous temperature residual at the next time point is used as output data to train the instantaneous temperature residual prediction model. The planned injected energy in each sampling period is calculated based on the command power, and the open-loop thermal instability degree is obtained based on the instantaneous temperature residual change characteristics in each sampling period. The command power, planned injected energy and open-loop thermal instability degree in the sampling period are used as input data, and the machine tool background temperature in the next sampling period is used as output data to train the machine tool background temperature prediction model. The instantaneous temperature residual prediction model is used to process the data under the real-time sampling period to obtain the predicted instantaneous temperature residual; the predicted machine tool background temperature residual is obtained based on the difference between the machine tool background temperature under the real-time sampling period and the predicted machine tool background temperature output by the machine tool background temperature prediction model. The laser power is adjusted based on the predicted instantaneous temperature residual, and the machine tool feed rate is adjusted based on the predicted instantaneous temperature residual and the predicted machine tool background temperature residual.

2. The intelligent thermal error compensation method for a numerical control system based on multi-source temperature sensing according to claim 1, characterized in that, The static heat dissipation characteristics include geometric heat dissipation conditions and path reheating risk; the volume ratio of solidified entity to powder to be melted in a local area of ​​each path point is obtained, and the average of all volume ratios is taken as the geometric heat dissipation conditions; the number of reheating times at each path point is counted, and the average of the number of reheating times at all path points is taken as the path reheating risk.

3. The intelligent thermal error compensation method for a numerical control system based on multi-source temperature sensing according to claim 1, characterized in that, The machine tool background temperature is the average value of all elements in the machine tool structure temperature feature vector.

4. The intelligent thermal error compensation method for a CNC system based on multi-source temperature sensing according to claim 1, characterized in that, The method for obtaining the processing temperature includes: The input temperature is obtained by multiplying the command power by a preset conversion coefficient, and the sum of the input temperature and the machine tool background temperature is taken as the processing temperature.

5. The intelligent thermal error compensation method for a numerical control system based on multi-source temperature sensing according to claim 1, characterized in that, The method for obtaining the instantaneous temperature residual includes: The difference between the instantaneous molten pool temperature and the processing temperature is taken as the instantaneous temperature residual.

6. The intelligent thermal error compensation method for a numerical control system based on multi-source temperature sensing according to claim 1, characterized in that, The degree of open-loop thermal instability is the standard deviation of the instantaneous temperature residuals at all time points within the sampling period.

7. The intelligent thermal error compensation method for a numerical control system based on multi-source temperature sensing according to claim 1, characterized in that, The method for obtaining the predicted machine tool background temperature includes: The instantaneous temperature residual prediction model is used to obtain the predicted instantaneous temperature residual sequence for the real-time sampling period. The predicted open-loop thermal instability degree is obtained based on the predicted instantaneous temperature residual sequence. The predicted open-loop thermal instability degree, the planned injection energy under the real-time sampling period, and the command power under the real-time sampling period are used as input data, and the predicted machine tool background temperature is output through the machine tool background temperature prediction model.

8. The intelligent thermal error compensation method for a numerical control system based on multi-source temperature sensing according to claim 1, characterized in that, The predicted machine tool background temperature residual is the difference between the predicted machine tool background temperature and the machine tool background temperature during the real-time sampling period.

9. The intelligent thermal error compensation method for a numerical control system based on multi-source temperature sensing according to claim 1, characterized in that, The adjustment of laser power based on the predicted instantaneous temperature residual includes: Divide the predicted instantaneous temperature residual by a preset conversion coefficient to obtain the compensation power; subtract the compensation power from the laser's base power to obtain the adjusted laser power.

10. The intelligent thermal error compensation method for a numerical control system based on multi-source temperature sensing according to claim 1, characterized in that, The adjustment of the machine tool feed rate based on the predicted instantaneous temperature residual and the predicted machine tool background temperature residual includes: The sum of the predicted instantaneous temperature residual and the predicted machine tool background temperature residual is used as the cumulative value of the machine tool thermal error factor. If the cumulative value of the machine tool thermal error factor is greater than a preset first threshold and less than a preset second threshold, the feed rate is reduced by a linear decreasing method. If the cumulative value of the machine tool thermal error factor is greater than or equal to the second threshold, the feed rate is adjusted to a preset minimum value.

Images