Motion precision compensation system and method of numerical control precision automatic lathe

Abstract

The invention discloses a motion precision compensation system and method of a numerical control precision automatic lathe, and aims to solve the problems that an existing precision compensation method depends on external calibration, production is interrupted, dynamic errors cannot be responded in real time, and the self-learning closed-loop capability is lacked. The method comprises the steps that servo current, spindle vibration and tool turret position signals are collected in real time; extracting signal features and inputting the signal features into a pre-training model to evaluate a rigidity degradation degree and predict a position deviation; querying a compensation parameter library according to an evaluation result; and injecting the compensation offset into a subsequent numerical control instruction in real time. The system comprises a multi-source signal acquisition module, an edge AI inference engine, a precision state evaluation model, a self-adaptive compensation parameter library and a CNC instruction injection interface. By the adoption of the technical scheme, online real-time precision sensing and self-adaptive compensation without external high-precision equipment and production interruption can be achieved, and machining size consistency and long-term stability are remarkably improved.

Description

TECHNICAL FIELD

[0001] The application belongs to the field of machine tool control, and particularly relates to a motion precision compensation system of a numerical control walking center machine. BACKGROUND

[0002] As core equipment for high-precision micro part machining, the numerical control walking center machine is widely used in the fields of medical devices, precision electronics and aerospace, etc. which have strict requirements on size consistency and surface integrity. The numerical control walking center machine realizes one-time forming machining of complex profiles through clamping workpieces by a main shaft and cooperating with a multi-axis linkage tool turret, and has the characteristics of high rigidity, high speed and high integration. In such a machine tool, the positioning accuracy and trajectory roundness of the X / Z axis directly determine the geometric quality of the final workpiece, and are key indicators for measuring the performance of the whole machine.

[0003] Among them, the motion precision compensation technology aims to offset systematic deviations caused by mechanical wear, thermal deformation and assembly gap evolution, etc. to maintain long-term machining stability. Existing compensation means mainly includes two categories: one is offline calibration compensation, which relies on external high-precision instruments such as laser interferometer or ball bar to establish a static error model in the stopped state, which has high precision but interrupts the production process and cannot capture real-time degradation in dynamic working conditions; the other is closed-loop feedback control, which uses grating or encoder to form a full-closed-loop servo system, which can only suppress the following error within the servo bandwidth, and is difficult to cope with large-scale position deviation caused by structural stiffness reduction due to guide rail wear and main shaft loosening.

[0004] The existing technology further exposes multiple limitations: the prediction method based on machine learning generally lacks the ability to deeply mine the internal signals of the machining process, and fails to convert existing information such as servo current harmonics and main shaft vibration spectrum into precision state perception basis; more importantly, the current system generally lacks a closed-loop mechanism of "self-diagnosis - self-decision - self-execution", and cannot realize online update and feedforward injection of compensation parameters without human intervention. The above problems lead to the dilemma of response lag, poor adaptability and high maintenance cost of existing compensation strategies in continuous batch production scenarios, and an motion precision compensation system that is free of external calibration, relies on original machine tool signals and has real-time self-learning ability is needed to support long-term maintenance of sub-micron level machining stability. SUMMARY

[0005] The purpose of the present application is to make up for the shortcomings of the prior art, and to provide an motion precision compensation system of a numerical control walking center machine, which can effectively solve the problems in the background art.

[0006] To achieve the above purpose, the present application provides the following technical scheme: On the one hand, an motion precision compensation system of a numerical control walking center machine is proposed, which includes the following components: A multi-source signal acquisition module is configured to acquire multi-source signals including a servo current signal, a spindle vibration signal and a turret position signal in a machine tool in real time during a machining process, and the multi-source signals are used as internal process variables representing a motion accuracy state; An edge AI inference engine is deployed in an edge computing unit of the machine tool, and is configured to pre-process the multi-source signals and extract time-frequency domain features, the time-frequency domain features including a total harmonic distortion rate of the servo current, an energy ratio of the spindle vibration in a specific frequency band, and a position back-lash of the turret; An accuracy state evaluation model is a pre-trained lightweight neural network model, and the input of the model is a feature vector extracted by the edge AI inference engine, and the output of the model is a stiffness index representing a current comprehensive stiffness degradation degree of the machine tool and a predicted X / Z axis position deviation vector; An adaptive compensation parameter library is configured to store a mapping relationship between different machining condition parameters and corresponding compensation offsets, and the machining condition parameters at least include a workpiece material, a spindle speed and a feed rate; A CNC instruction injection interface is configured to superimpose the compensation offset output by the adaptive compensation parameter library into a CNC program interpolation instruction stream of a subsequent workpiece in real time through a standard application programming interface provided by a numerical control system, so as to realize a feed-forward accuracy compensation.

[0007] Preferably, the multi-source signal acquisition module multiplexes or supplements original sensors of the machine tool. The servo current signal is acquired through high-precision Hall current sensors connected to output ends of X-axis and Z-axis servo drivers, and a sampling frequency is not less than 20 kHz. The spindle vibration signal is acquired through a three-axis MEMS vibration sensor installed on a front flange of the spindle, and a sampling frequency is not less than 5 kHz. The turret position signal is provided by an original absolute value encoder of the machine tool through a field bus protocol.

[0008] Further, the feature extraction process performed by the edge AI inference engine is specifically as follows: performing a fast Fourier transform on the servo current signal to calculate a total harmonic distortion rate (THD), and a calculation formula of the total harmonic distortion rate is , wherein is an effective value of a fundamental current, is an effective value of an hth harmonic current; performing a short-time Fourier transform on the spindle vibration signal to extract a percentage of vibration energy in a 100 Hz to 500 Hz frequency band in total energy; and recording a difference between a final position and an instruction target position of the encoder when each tool changing action of the turret is completed as a position back-lash.

[0009] In addition, the precision state evaluation model adopts a one-dimensional convolutional neural network structure, an input layer of which receives a three-dimensional feature vector composed of THD, vibration energy proportion, and back difference amount. The model contains two convolutional layers, one global average pooling layer, and two fully connected layers, and the final output layer is divided into two branches: one branch outputs a stiffness index scalar with a range of 0 to 1, and the other branch outputs a two-dimensional position deviation vector containing Δx and Δz components. The model is quantized to FP16 precision by the TensorRT framework before deployment to ensure that the single inference delay on the edge computing unit is less than 8 ms. The embodiment preferably adopts a one-dimensional convolutional neural network, but those skilled in the art can understand that, under the premise of meeting the computing power and delay requirements of the edge computing unit, other lightweight model architectures, such as a simplified version of a recurrent neural network (RNN) or a Transformer, can also be used to implement the functions of the present application.

[0010] In another aspect, a motion precision compensation method for a numerical control turret lathe is also proposed, and the specific steps are as follows: Step S110, in the process of machining a workpiece by the numerical control turret lathe, three-phase current signals of X-axis and Z-axis servo motors, vibration acceleration signals of the main shaft in three orthogonal directions, and rotation to position feedback signals of the tool turret are synchronously and in real time collected; Step S120, the collected current signals and vibration signals are respectively preprocessed by denoising and standardization, and the total harmonic distortion rate of the servo current, the energy proportion of the main shaft vibration signal in the 100-500 Hz frequency band, and the position back difference amount of the tool turret during tool changing are extracted to form a feature vector; Step S130, the feature vector is input into a pre-trained lightweight convolutional neural network model, and the model synchronously calculates and outputs a stiffness index representing the current motion precision degradation degree of the machine tool and a predicted X / Z-axis position deviation vector; Step S140, it is judged whether the stiffness index is lower than a preset degradation threshold; if yes, according to the current machining condition parameters including the workpiece material type, the main shaft speed and the feed rate, the adaptive compensation parameter library is queried to search for a matched compensation offset; if no, the existing compensation parameter is maintained or set to zero; Step S150, through a standard API interface provided by a numerical control system controller, the compensation offset is injected into the starting point instruction of the numerical control program G code interpolation path of the next workpiece to be machined in a vector superposition manner in real time; Step S160, steps S110 to S150 are cyclically executed, so that the machining process, precision state perception, and compensation parameter adjustment form a closed loop to realize adaptive precision maintenance.

[0011] Compared with the prior art, the present application has the following beneficial effects: 1. Completely abandon the dependence on external high-precision measuring equipment such as laser interferometer, ball bar, etc., through deep mining and using the original or low-cost supplemented built-in sensor signals of the machine tool, the machining process itself is converted into a high-dimensional precision state perception source, which significantly reduces the system hardware cost and integration maintenance complexity.

[0012] 2. Real-time and continuity of compensation process is realized. The system can complete a complete signal acquisition, feature extraction, state evaluation and compensation decision in each workpiece machining period, and the compensation parameters can be dynamically updated with the machining process, effectively overcoming the disadvantages of interrupting production and unable to respond to real-time degradation in traditional offline calibration compensation.

[0013] 3. Strong self-learning and adaptive ability. Through the "working condition-signal-precision" data accumulated by long-term operation, the precision state evaluation model can be retrained and optimized, and the mapping relationship in the adaptive compensation parameter library can be dynamically updated, so as to continuously improve the long-term compensation precision and stability of the system under different wear stages and machining tasks.

[0014] 4. Strong system compatibility and engineering practicability. The CNC instruction injection interface is developed based on the standard communication protocols and APIs provided by mainstream numerical control systems such as Fanuc FOCAS, Siemens OPC UA, Mitsubishi MELSEC, etc., without deep modification of the original control kernel of the machine tool, which is convenient for modification and deployment on existing equipment. BRIEF DESCRIPTION OF DRAWINGS

[0015] Figure 1 is the overall technical scheme architecture schematic diagram of the motion precision compensation system of the numerical control walking heart machine proposed by the present application; Figure 2 is the precision state evaluation and adaptive compensation closed-loop logic flow framework diagram based on multi-source signal perception and AI inference in the present application; DETAILED DESCRIPTION

[0016] The technical solutions of the present application will be described in detail below in conjunction with the embodiments. Obviously, the described embodiments are only a part of the embodiments of the present application, not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0017] In order to further illustrate the technical means and effects adopted by the present application to achieve the predetermined invention purpose, the specific embodiments according to the present application are described in detail as follows in combination with the drawings and preferred embodiments.

[0018] Embodiment one

[0019] In high-volume, long-term continuous machining scenarios for precision shaft parts, such as the manufacturing of fuel injector valve bodies for automotive engines, CNC Swiss-type lathes need to maintain extremely high dimensional accuracy and shape tolerance consistency. Traditional methods relying on external equipment such as laser interferometers for periodic offline calibration not only interrupt production but also fail to capture real-time accuracy degradation caused by thermal deformation and mechanical wear during machining. This embodiment provides a motion accuracy compensation system for CNC Swiss-type lathes, aiming to achieve adaptive accuracy maintenance during the machining process.

[0020] See Figure 1 This system is physically integrated into the CNC Swiss-type lathe body and its CNC system. Its core components include a multi-source signal acquisition module, an edge AI inference engine, a precision status assessment model, an adaptive compensation parameter library, and a CNC instruction injection interface. These modules work together to form a closed loop from signal perception to compensation execution.

[0021] The multi-source signal acquisition module is responsible for continuously capturing process variables characterizing the intrinsic state of machine tool motion accuracy during machining. This module fully utilizes or adds to the machine tool's existing sensors at low cost, avoiding the introduction of complex external measuring devices. Specifically, for servo current signal acquisition, a non-invasive closed-loop Hall current sensor is connected in parallel to the three-phase current output terminals of the X-axis and Z-axis servo drives necessary for high-precision machining. This sensor has a bandwidth of at least 100 kHz and a sampling frequency set to 20 kHz to ensure complete capture of high-frequency harmonic components in the current waveform. The current signal is transmitted to the signal conditioning circuit via shielded twisted-pair cable for preliminary amplification and filtering, and then converted into a digital signal stream by a 16-bit analog-to-digital converter. For spindle vibration signal acquisition, a three-axis microelectromechanical system (MEMS) vibration sensor is rigidly mounted on the non-rotating reference surface of the spindle's front flange. This sensor measures the spindle's vibration acceleration in three orthogonal directions: radial, axial, and tangential, with a range of ±50 g and a sampling frequency set to 5 kHz. The sensor output is transmitted directly via a digital interface after passing through a built-in anti-aliasing filter. For acquiring the turret position signal, the original multi-turn absolute encoder of the machine tool is directly reused. This encoder periodically reports the real-time rotation angle of the turret and the final position after the tool change to the CNC system in the form of digital messages via PROFIBUS-DP or MELSECNET fieldbus protocol. The multi-source signal acquisition module uses a dedicated field-programmable gate array to synchronously acquire and timestamp multiple signals, ensuring that all signal samples have strict time synchronization, laying the foundation for subsequent time-frequency domain correlation analysis.

[0022] The edge AI inference engine is deployed on an industrial-grade edge computing unit directly installed in the electrical cabinet. The unit is equipped with a multi-core processor and a dedicated neural network inference accelerator. The primary task of the engine is to preprocess and extract features from the raw signal streams. For the servo current digital signal stream, the engine first applies a sliding window-based Kalman filter for denoising to suppress random noise caused by electromagnetic interference. Subsequently, a fast Fourier transform is performed on a data window consisting of 1024 sample points to calculate the current frequency spectrum. Based on the spectrum data, the engine performs core feature calculation: total harmonic distortion rate. The calculation formula is , where is the fundamental current effective value, is the hth harmonic current effective value, and H is 50 to cover the main harmonic components. The total harmonic distortion rate directly reflects the load characteristic distortion of the servo drive system caused by changes in mechanical friction, relaxation of guide rail pre-tightening force. For the spindle vibration acceleration signal, the engine first performs coordinate transformation to combine the three-axis signals into a total vibration vector. Then, a Hamming window is applied to a data segment consisting of 512 sample points, and a short-time Fourier transform is performed to generate a time-frequency spectrum. The engine extracts the vibration energy in the key frequency band of 100 Hz to 500 Hz and calculates its percentage of the total vibration energy. This frequency band energy percentage is a sensitive indicator for identifying spindle bearing wear and dynamic balance degradation. For the tool turret position signal, the engine continuously monitors the position feedback of the encoder. Whenever a tool change command is detected to be executed and the position feedback is stable, the difference between the absolute position value of the encoder feedback and the target command position value issued by the numerical control system is recorded, which is defined as the position back-lash of this tool change. The cumulative position back-lash reflects the conditions of gear clearance and worm wear in the tool turret transmission chain. Finally, the edge AI inference engine combines the total harmonic distortion rate, vibration energy percentage, and the latest position back-lash calculated in real time into a 3-dimensional feature vector and sends it to the downstream precision state evaluation model.

[0023] The precision state evaluation model is a lightweight one-dimensional convolutional neural network that is pre-trained offline and deployed. Before deployment, the model is trained using a large amount of "feature vector-actual precision deviation" data collected from multiple machines of the same type during actual machining for several months. The actual precision deviation data is obtained by measuring the key dimensions of a standard test piece using a contact online probe before the machine is shipped or during regular maintenance, and the difference between the measurement result and the theoretical size is used as the label data. During the long-term operation of the system, the non-contact laser range finder integrated in the tool holder can also be used to quickly scan the key features of the workpiece during the tool changing gap to obtain precision feedback data for online updating of the model. After training, the model is quantized to FP16 precision by the TensorRT framework and optimized into an efficient inference engine suitable for edge computing units, ensuring that the end-to-end delay of a single forward propagation is strictly controlled within 8 milliseconds. The network structure of the model includes 2 one-dimensional convolutional layers, 1 global average pooling layer, and 2 fully connected layers. The input layer receives the above-mentioned 3-dimensional feature vector. The forward propagation process of the model completes the nonlinear mapping and abstraction of high-dimensional features. The final output layer of the model is divided into two independent branches: the first branch outputs a scalar ranging from 0 to 1, defined as the "stiffness index". This index comprehensively represents the degree of rigidity degradation of the current mechanical structure of the machine tool (such as the bed, spindle box, and slide), and the lower the value, the more serious the rigidity loss. The second branch outputs a 2-dimensional vector containing Δx and Δz components, i.e., the predicted comprehensive positional deviation of the X-axis and Z-axis in the next machining cycle, with units of microns. This predicted deviation integrates the effects of thermal error, force-induced deformation, and geometric error.

[0024] The adaptive compensation parameter library is a relational database stored in the solid state disk of the edge computing unit. It establishes a nonlinear mapping relationship between "machining working condition parameters" and "ideal compensation offset". Among them, the machining working condition parameters are a multi-dimensional vector, at least including: workpiece material type (such as stainless steel, aluminum alloy, titanium alloy), spindle speed (unit: revolutions per minute), X-axis and Z-axis feed rate (unit: millimeters per minute). The compensation offset is a corresponding 2-dimensional vector (Cx, Cz), representing the instruction offset that needs to be applied in the corresponding axis to offset the predicted deviation. The initial data of the parameter library comes from the precision calibration experiment and finite element simulation when the machine is shipped. During the long-term operation of the system, the parameter library has the ability to update itself. After the system performs compensation, the compensation effect can be verified in reverse through online measurement results (such as contact probes) after machining or regular inspection of workpiece dimensions, and the mapping entries in the library are corrected or added to realize the accumulation and optimization of experience.

[0025] The CNC instruction injection interface is a key execution component for implementing compensation actions. This interface software is developed based on standard application programming interfaces (APIs) provided by CNC system manufacturers, such as the FOCAS library for Fanuc systems, the OPC UA client for Siemens systems, and the MELSEC communication protocol for Mitsubishi systems. This interface runs as a foreground or background service, continuously monitoring the compensation offset output from the adaptive compensation parameter library. When compensation is needed, before the CNC program for the next workpiece begins execution, the interface precisely adds the compensation offset (Cx, Cz) as a coordinate offset to the current workpiece coordinate system of the CNC system via API function calls, at the starting point of the interpolator's G-code path planning. This process is a feedforward superposition, which does not modify the original CNC program file, only affecting the real-time interpreted instruction flow, thus achieving immediate correction of subsequent machining trajectories.

[0026] See Figure 2 The above modules work together in strict timing and logical closed loop, specifically as shown in the following method steps: In step S110, during the machining of a single workpiece by a CNC Swiss-type lathe, the system synchronously initiates multi-source signal acquisition. The three-phase current signals of the X-axis and Z-axis servo motors, the three-axis vibration acceleration signals of the spindle, and the rotational position feedback signal of the turret are acquired and buffered in real time and synchronously.

[0027] In step S120, the edge AI inference engine performs batch processing on the cached current and vibration signal data. First, denoising and standardization preprocessing is performed to eliminate the influence of dimensions. Then, the engine strictly follows the aforementioned method to extract the total harmonic distortion rate of the servo current, the energy proportion of the spindle vibration in the 100–500 Hz frequency band, and calculates the position backlash generated by the most recent turret tool change within the current machining cycle. These three feature values ​​are constructed into a standardized 3D feature vector.

[0028] In step S130, the feature vector is immediately fed into the accuracy status assessment model that has been loaded into memory for inference. The model performs parallel computation and outputs two results simultaneously: a stiffness index (scalar) characterizing the degree of degradation of the machine tool's current overall motion accuracy, and a predicted X / Z axis position deviation vector (Δx, Δz).

[0029] Step S140, the system decision logic is triggered. The core of the decision is to determine whether the stiffness index is lower than the preset degradation threshold (for example, 0.75). If the stiffness index is not lower than the threshold, it is considered that the current state of the machine tool is good, and the prediction deviation may be dominated by random factors, and the system will maintain the existing compensation parameters unchanged, or the compensation offset will be set to zero. If the stiffness index is lower than the threshold, it indicates that the machine tool has significant rigidity degradation, and the prediction deviation has systematicity. At this time, the system obtains the current machining condition parameters, including the type of workpiece being machined, the current spindle speed program value, and the current feed rate program value. With these working condition parameters as the index, the adaptive compensation parameter library is queried. The query algorithm adopts a combination of nearest neighbor matching and linear interpolation, retrieves one or more sets of compensation offsets from the library that best match the current working conditions, and finally determines the compensation offset (Cx, Cz) to be used this time through weighted calculation.

[0030] Step S150, after the compensation offset is determined, the CNC instruction injection interface immediately acts. At the moment when the current workpiece machining is completed and the numerical control program segment of the next workpiece is about to be executed by the interpolator, the interface injects (Cx, Cz) as the coordinate offset into the path planning buffer of the numerical control system through the standard API of the numerical control system. The offset will take effect from the next interpolation period, and real-time corrects the tool movement trajectory.

[0031] Step S160, after completing the machining and compensation of a workpiece, the system immediately resets and executes steps S110 to S150 in a loop. The machining period of each workpiece (usually tens of seconds to minutes) constitutes a complete "perception-evaluation-decision-execution" closed loop. Through this high-frequency closed loop adjustment, the machining process, precision state perception and compensation parameter adjustment are closely coupled, so that the system can dynamically track and compensate the precision drift caused by thermal accumulation and progressive wear, and realize adaptive precision maintenance.

[0032] The system described in the embodiment has been verified by actual installation. After 8 hours of continuous machining, the key shaft diameter size consistency of the workpiece machined by the numerical control lathe using the system can be improved from ±5 microns to within ±1.2 microns; in a 72-hour unattended batch machining, the process capability index CPK of the key hole diameter is significantly improved from 0.92 to 1.67, realizing long-term maintenance of sub-micron machining stability, and greatly reducing the scrap rate and quality inspection cost.

[0033] Embodiment two

[0034] In the mass production of precision stainless steel bone pins in the medical device field, higher requirements are put forward for the dynamic accuracy of the numerical control turning machine and the surface machining quality. Based on the system architecture of embodiment one, this embodiment further strengthens the monitoring and compensation of the thermal state characteristics of the spindle and the chatter in the cutting process, and refines the dimensions of the compensation parameter library.

[0035] In this embodiment, the multi-source signal acquisition module adds one signal acquisition based on embodiment one: through a non-contact infrared temperature sensor, the temperature of the outer ring of the spindle front bearing is monitored in real time, and the sampling frequency is 1 Hz. This temperature signal is used to more accurately model the influence of spindle thermal elongation on Z-axis direction accuracy. At the same time, the frequency band analysis of the spindle vibration signal is expanded, in addition to the 100-500 Hz wear frequency band, the 500-2000 Hz high frequency band is additionally concerned, and the sudden increase of vibration energy in this frequency band is an important precursor feature of the impending cutting chatter.

[0036] The feature extraction process of the edge AI inference engine is correspondingly expanded. For the temperature signal, calculate its slope relative to the ambient temperature rise. For the vibration signal, in addition to calculating the energy ratio of the original concerned frequency band, a new high frequency band (500-2000 Hz) energy and low frequency band (0-500 Hz) energy ratio is calculated as the "chatter tendency index". Therefore, the feature vector dimension of this embodiment is expanded from 3D to 5D, including: servo current total harmonic distortion rate, spindle vibration low frequency energy ratio, spindle vibration high frequency / low frequency energy ratio, spindle bearing temperature rise slope, cutter tower position return difference.

[0037] The precision state evaluation model also needs to be retrained to adapt to 5D input. Its output, in addition to retaining the stiffness index and X / Z-axis basic position deviation vector, adds a "surface quality attenuation coefficient" scalar output. This coefficient is used to predict the degree of deterioration of the workpiece surface roughness that may be caused by continuing machining under the current state.

[0038] The dimension of the adaptive compensation parameter library has increased significantly. In addition to the workpiece material, spindle speed, and feed rate, the cutting depth, cutting width, and cooling liquid state (on / off) are added. The compensation offset is also expanded to a 4D vector (Cx, Cz, Cs, Cf), where Cx and Cz are position offsets, Cs is the spindle speed adjustment amount (unit: revolutions per minute) for surface quality attenuation, and Cf is the feed rate adjustment percentage for chatter tendency. The construction of the parameter library is based on a large number of cutting experiments and surface roughness detection data.

[0039] The function of the CNC instruction injection interface is enhanced. In addition to being able to inject coordinate bias, it also needs to have the ability to dynamically fine-tune the spindle speed S instruction and the feed rate F instruction through the API. When the system output compensation vector contains Cs and Cf, the interface needs to temporarily cover the corresponding mode instruction in the numerical control program for temporary modification during the compensation period.

[0040] The method steps are refined based on Embodiment One: In step S110, the synchronously collected signals add spindle bearing temperature.

[0041] In step S120, feature extraction adds the calculation of temperature rise slope and chatter tendency index, forming a 5-dimensional feature vector.

[0042] In step S130, the precision state evaluation model outputs the stiffness index, the position deviation vector (Δx, Δz), and the surface quality attenuation coefficient.

[0043] In step S140, the decision logic becomes a multi-condition judgment. First, determine whether either the stiffness index or the surface quality attenuation coefficient is below its corresponding threshold. Second, determine whether the chatter tendency index exceeds the safety threshold. According to different threshold trigger combinations, query the compensation parameter library. For example, if only the stiffness index is low, mainly query the position compensation amount (Cx, Cz); if the chatter tendency index is high, prioritize querying the feed rate adjustment amount Cf; if the surface quality attenuation coefficient is low and the temperature rise is high, it may query the compound compensation amount containing spindle speed fine tuning Cs.

[0044] In step S150, the CNC instruction injection interface performs compound compensation actions according to the obtained 4-dimensional compensation vector (Cx, Cz, Cs, Cf): injects coordinate bias (Cx, Cz), and temporarily covers the spindle speed (S instruction increases Cs) and feed rate (F instruction multiplied by (1+Cf)) through system instructions.

[0045] Through the enhancement of this embodiment, the system not only compensates for geometric position errors, but also actively suppresses chatter and optimizes cutting parameters to maintain surface quality, realizing an upgrade from "precision compensation" to "process optimization", especially suitable for medical device part machining with extremely high surface integrity requirements.

[0046] The specific embodiments of the application have been described above. It should be understood that the application is not limited to the above specific embodiments, and those skilled in the art can make various changes or modifications within the scope of the claims, which does not affect the essential content of the application. In the case of no conflict, the embodiments of the present application and the features in the embodiments can be arbitrarily combined with each other.

Claims

1. A motion accuracy compensation system for a numerical control lathe, characterized by, The system comprises the following components: A multi-source signal acquisition module for acquiring multi-source signals including servo current signals, spindle vibration signals and turret position signals built-in the machine tool in real time during the machining process, the multi-source signals serving as intrinsic process variables representing the motion accuracy state; An edge AI inference engine deployed in a local edge computing unit of the machine tool for pre-processing the multi-source signals and extracting time-frequency domain features, the time-frequency domain features including total harmonic distortion rate of the servo current, energy proportion of the spindle vibration in a specific frequency band and turret position back-lash amount; An accuracy state evaluation model, which is a pre-trained lightweight neural network model, takes the feature vector extracted by the edge AI inference engine as input and outputs a stiffness index representing the current comprehensive stiffness degradation degree of the machine tool and a predicted X / Z axis position deviation vector; An adaptive compensation parameter library storing the mapping relationship between different machining condition parameters and corresponding compensation offsets, the machining condition parameters including at least workpiece material, spindle speed and feed rate; A CNC instruction injection interface for superimposing the compensation offset output by the adaptive compensation parameter library into the subsequent workpiece's CNC program interpolation instruction stream in real time through the standard application programming interface provided by the numerical control system, so as to realize feed-forward accuracy compensation.

2. The motion accuracy compensation system of a CNC lathe according to claim 1, wherein, In the multi-source signal acquisition module, the servo current signals are acquired by connecting high-precision Hall current sensors to the output terminals of the X-axis and Z-axis servo drivers, with a sampling frequency not lower than 20 kHz; the spindle vibration signals are acquired by a three-axis MEMS vibration sensor installed on the front flange of the spindle, with a sampling frequency not lower than 5 kHz; and the turret position signals are provided by the original absolute value encoder of the machine tool through the field bus protocol.

3. The motion accuracy compensation system of a CNC lathe according to claim 1, wherein, The feature extraction process performed by the edge AI inference engine is specifically: performing fast Fourier transform on the servo current signal, calculating the total harmonic distortion rate thereof, and the calculation formula is , wherein is the effective value of the fundamental wave current, is the effective value of the hth harmonic current; performing short-time Fourier transform on the main shaft vibration signal, and extracting the percentage of vibration energy in the 100 Hz to 500 Hz frequency band in the total energy; recording the difference between the final position and the instruction target position of the encoder feedback when each tool changing action of the tool tower is completed as the position back difference.

4. The motion accuracy compensation system of a CNC lathe according to claim 1, wherein, The accuracy state evaluation model adopts a one-dimensional convolutional neural network structure, with the input layer receiving a 3-dimensional feature vector composed of total harmonic distortion rate, vibration energy proportion and back-lash amount; the model includes two convolutional layers, one global average pooling layer and two fully connected layers, and the final output layer is divided into two branches: one branch outputs a stiffness index scalar with a range of 0 to 1, and the other branch outputs a 2-dimensional position deviation vector containing Δx and Δz components; the accuracy state evaluation model is quantized to FP16 precision before deployment to ensure that the single inference delay on the edge computing unit is less than 8 ms.

5. The motion accuracy compensation system of a CNC lathe according to claim 1, wherein, In the adaptive compensation parameter library, the machining condition parameters also include cutting depth and cutting width; and the compensation offset is expanded to a 4-dimensional vector containing X-axis compensation offset Cx, Z-axis compensation offset Cz, spindle speed fine adjustment amount Cs and feed rate fine adjustment percentage Cf; The mapping relationship of the parameter library is constructed based on cutting experiments and surface roughness detection data, and has the ability to be self-updated according to online measurement results.

6. The motion accuracy compensation system of a CNC lathe according to claim 1, wherein, The CNC instruction injection interface is developed based on Fanuc FOCAS, Siemens OPC UA or Mitsubishi MELSEC standard communication protocol and API; the interface superimposes the compensation offset in the form of coordinate bias into the current workpiece coordinate system of the numerical control system before the next to-be-processed workpiece numerical control program starts to execute, and can dynamically fine-tune the spindle speed instruction and the feed rate instruction through API according to the Cs and Cf components in the compensation offset.

7. A motion accuracy compensation method for a numerical control wheel-cutting machine based on the motion accuracy compensation system according to any one of claims 1 to 6, characterized by, The method comprises the following steps: In step S110, the signals collected synchronously further include a spindle front bearing outer ring temperature signal obtained by a non-contact infrared temperature sensor; in step S120, the feature extraction further includes calculating a temperature rise slope of the spindle bearing and calculating an energy ratio of the spindle vibration signal in a 500-2000 Hz high frequency band and a 0-500 Hz medium-low frequency band as a chattering tendency index, and the feature vector is expanded to a 5-dimensional vector containing the total harmonic distortion rate, the vibration low-frequency energy ratio, the vibration high-frequency / low-frequency energy ratio, the temperature rise slope and the position back-off amount. In step S130, the precision state evaluation model further synchronously outputs a surface quality attenuation coefficient; in step S140, the decision logic performs multi-condition judgment based on the comparison results of the stiffness index, the surface quality attenuation coefficient and the chattering tendency index with respective preset threshold values, and queries the adaptive compensation parameter library to obtain a corresponding 4-dimensional compensation offset (Cx, Cz, Cs, Cf) according to different threshold value trigger combinations. ​ ​ ​ ​ 8. The motion accuracy compensation method of the CNC wheel cutting machine according to claim 7, characterized in that, ​ 9. The method of claim 8, wherein, ​ 10. The method of claim 9, wherein, In step S150, the injection of the compensation offset includes: injecting Cx and Cz as coordinate offset amounts into the numerical control system path planning buffer, and temporarily covering and modifying the spindle speed instruction by increasing Cs through a system instruction, and multiplying the feed rate instruction by (1+Cf).

Images