Change gear assembly machining process parameter intelligent optimization system

Abstract

The invention relates to the technical field of industrial automation control, and discloses a change gear assembly processing technological parameter intelligent optimization system, which comprises a multi-dimensional data synchronous acquisition unit, an angular domain static map construction unit, a phase lead constraint control unit and an angle trigger execution unit, according to the method, a static angle-load topological graph with a main shaft angle as an index is established, a phase lead angle is calculated by combining inherent physical response lag time of a servo system, and a parameter adjustment instruction sequence is directly injected into a servo current loop at a preset angle window before physical load impact occurs; according to the feed-forward control mechanism based on angular domain reconstruction and phase lead constraint, the problem of physical lag of traditional time domain feedback control in response to high-frequency intermittent cutting is avoided, and zero time difference of periodic load impact and even pre-judgment torque compensation are achieved.

Description

TECHNICAL FIELD

[0001] The present application relates to a hanging wheel assembly machining process parameter intelligent optimization system, belonging to the field of industrial automation control technology. BACKGROUND

[0002] In current hanging wheel assembly numerical control machining, maintaining cutting load balance to ensure part quality and tool life, the existing process parameter control generally adopts a real-time load feedback adaptive mode based on the comparison of spindle load and threshold value, adjusts the feed ratio to stabilize the cutting power, and the control logic is widely used in continuous and smooth cutting conditions. Hanging wheel machining presents typical high-frequency intermittent cutting characteristics, and the load change has step and periodicity. Under this condition, the time domain feedback strategy faces the challenge of physical response lag. Limited by signal filtering processing, numerical control system interpolation period and servo motor mechanical inertia, there is a phase delay from sensing load mutation to the completion of speed adjustment by the actuator. Under the high-frequency impact of intermittent cutting, the phase lag leads to time sequence mismatch between control action and actual cutting state.

[0003] In order to overcome the physical limitations of hardware response, digital means parameter optimization attempts have appeared in the industry, for example, the Chinese invention patent with publication number CN120542897A discloses a production line process parameter automatic optimization method based on digital twin technology. This scheme establishes a production line rotation model and multiple weight simulation to solve the problem of transmission efficiency decline caused by equipment aging. This kind of scheme belongs to global parameter optimization based on quasi-static model. The simulation correction logic focuses on long-period macro efficiency matching and trend adjustment. The calculation and response rhythm are slower than the transient changes in the cutting process. Facing the millisecond-level load step of hanging wheel machining, the lack of phase-advance prediction mechanism optimization means cannot break through the inherent mechanical and electrical inertia bottleneck of the servo system, and it is difficult to realize zero-time difference torque compensation within the extremely short window of tool cutting.

[0004] Therefore, how to avoid the time domain feedback limitation in the existing physical lag control system and realize precise compensation of high-frequency intermittent cutting impact has become a technical problem to be solved by the present application. SUMMARY

[0005] To solve the problems raised in the background art, the technical solution of the present application is as follows: a hanging wheel assembly machining process parameter intelligent optimization system, comprising: a multi-dimensional data synchronous acquisition unit, connected to a servo driver and a position encoder, using a hardware interrupt signal to trigger a parallel latching logic, and establishing a hard synchronization mapping interface of spindle load current time sequence and spindle angle position sequence.

[0006] An angular domain static atlas construction unit performs dimension transformation from time domain to angular domain, generates a static angle-load topology atlas with spindle angle as the only index and load current characteristic value as the mapping object through phase alignment and statistical filtering of multi-period synchronous data, and the atlas is used to eliminate the time dimension variable of servo control.

[0007] The phase lead constraint control unit calculates the phase lead angle based on the product of the inherent physical response lag time of the servo system and the real-time angular velocity of the main shaft, and generates a parameter adjustment instruction sequence with a negative time axis offset attribute according to the distribution position of the load mutation characteristic area in the static angle-load topological atlas.

[0008] The angle trigger execution unit locks the current angular position of the main shaft in real time during the control operation, and directly injects the parameter adjustment instruction sequence into the servo current loop within a preset angular window before the physical load impact occurs under the control of the phase lead angle, so as to realize zero-time difference torque compensation for the periodic discontinuous cutting load at the physical level.

[0009] Preferably, the angle domain static atlas construction unit further includes atlas dynamic migration logic for processing individual differences of workpieces, including: a feature anchor point capture module that monitors the first derivative of the load current in the first waveform of the control period, locks the instantaneous phase angle of the actual waveform rising edge and the actual peak load; a deviation operator calculation module that calculates the phase offset between the instantaneous phase angle and the reference phase angle in the static angle-load topological atlas, and calculates the amplitude proportional coefficient between the actual peak load and the reference peak load; and an execution atlas correction module that performs a translation transformation on the index axis of the static angle-load topological atlas using the phase offset, and performs a linear gain scaling on the atlas load value using the amplitude proportional coefficient, to generate a temporary execution atlas adapted to the current control object.

[0010] Preferably, the angle domain static atlas construction unit further includes a zero clearance window dynamic zeroing logic for eliminating environmental thermal drift interference, including: a window locking module that identifies a non-working angular interval in the static angle-load topological atlas where the load current and its rate of change are both below a preset threshold, and marks the interval as a zero point calibration window; a dynamic bias update module that collects real-time load current and calculates a direct current component drift value when the main shaft rotates to the zero point calibration window; and a real-time correction module that subtracts the direct current component drift value calculated the last time from the real-time collected load current signal when the main shaft rotates to the working angular interval, and outputs the corrected load feedback data.

[0011] Preferably, the system further includes a virtual stiffness compensation unit for offsetting mechanical elastic deformation errors, which includes: a stiffness model storage module that stores equivalent stiffness coefficients of the controlled object; a deformation amount estimation module that calculates the expected elastic deformation amount of the controlled object at the current angle in real time according to the load current value in the static angle-load topological atlas and the equivalent stiffness coefficients; and a superimposed control injection module that generates a micro-displacement compensation instruction with an amplitude equal to and a direction opposite to the expected elastic deformation amount, and directly superimposes the instruction to the position loop control node of the feed shaft, independent of the speed loop control instruction to maintain the geometric position accuracy of the system.

[0012] Preferably, the phase lead constraint control unit further comprises a gain adaptive evolution module for compensating the resistance variation caused by tool wear, which comprises: a residual extraction submodule for calculating a control residual sequence between the measured load current and the preset target load value in real time within each machining period; an energy integration submodule for performing integral operation on the control residual sequence within the cutting angle interval to generate a period residual energy index representing the degree of mismatch of the feedforward model; and a gain scheduling submodule for updating the global feedforward gain coefficient based on the period residual energy index, which is used by the phase lead constraint control unit to dynamically scale the amplitude of the parameter adjustment instruction sequence.

[0013] Preferably, the gain scheduling submodule adopts a residual energy-based cumulative step algorithm to update the global feedforward gain coefficient, and the update rule followed is: wherein, is the updated global feedforward gain coefficient, is the global feedforward gain coefficient of the current period, is a preset learning rate step, is the period residual energy index, which is obtained by integrating the absolute value of the control residual sequence within the cutting interval.

[0014] Preferably, the phase lead angle in the phase lead constraint control unit is determined based on the product of the sum of the current loop filter delay time of the servo system, the calculation interpolation period, and the inertia lag constant of the mechanical transmission chain, and the real-time angular velocity of the spindle, and the phase lead angle defines the advance of the control instruction relative to the physical load occurrence position in the angle domain coordinate system.

[0015] Preferably, the multi-dimensional data synchronous acquisition unit adopts a field programmable gate array (FPGA) as the hardware carrier, and triggers the parallel latching of data through a hardware interrupt signal to ensure that the time deviation between the load current sampling time and the position encoder counting time is less than one-tenth of the servo control period.

[0016] Preferably, the parameter adjustment instruction sequence includes a feed rate adjustment instruction, and the phase lead constraint control unit is configured with inverse mapping logic that establishes an inverse proportional gain function between the feed rate and the cutting load, generates a low feed rate instruction in the high load characteristic region of the static angle-load topological map, and generates a high feed rate instruction in the low load characteristic region.

[0017] Preferably, the angle domain static map construction unit comprises a period superposition submodule for obtaining original cutting data streams of consecutive multiple rotation periods, and performing superposition average processing on the load data of the same angle index position using a synchronous average algorithm to eliminate non-periodic random noise interference.

[0018] Compared with the prior art, the beneficial effects of the present application are: 1. In the machining process of the hanging wheel assembly, the angle position of the main shaft and the load current sequence are synchronously collected, a load mapping atlas indexed by angle is constructed, which is used as a time reference for feedforward control, the periodic characteristics of the hanging wheel machining are utilized to convert the time domain load fluctuation into the angle domain spatial characteristics, the phase lead angle is calculated by combining the inherent response lag time of the servo system, the feed adjustment instruction is issued before the physical moment of load mutation occurs, the predictive compensation of high-frequency intermittent cutting impact is realized, the response lag caused by signal filtering delay and servo mechanical inertia in traditional feedback control is eliminated, and the speed adjustment of the feed shaft is completed at the moment when the tool cuts into the workpiece.

[0019] 2. The non-cutting angle interval in the intermittent cutting process is utilized to establish a dynamic zero point calibration logic based on process characteristics, the idle stroke window in the load atlas is identified within each rotation period, and the current reference value under the window is collected, so that the thermal drift of the sensor and the amplification circuit is obtained in real time, and is deducted in subsequent cutting data processing. This self-calibration mechanism enables the control system to maintain the absolute stability of the cutting load judgment threshold without relying on constant temperature detection hardware or periodic shutdown calibration, and avoids the interference of environmental temperature changes on the control accuracy.

[0020] 3. For the problem of nonlinear growth of cutting resistance caused by tool wear, a gain adaptive evolution mechanism based on residual energy integration is introduced, the control residual between the actual load and the target load is calculated in real time, the energy integration of the residual sequence in the cutting interval is extracted as an evaluation index, which represents the mismatch degree between the current feedforward model and the actual physical object, and the global gain coefficient of the feedforward instruction is dynamically corrected based on the evaluation index, so that the control system can automatically perceive and compensate for the process system characteristic drift caused by tool aging, ensure the consistency of the machining quality throughout the tool life cycle, and does not need manual adjustment of the control parameters; an inverse model of the process system elastic modulus is established, the current signal in the angle domain load atlas is converted into elastic deformation, a micro displacement compensation instruction opposite to the deformation direction is generated, and the instruction is superimposed on the feed shaft position ring control flow, which is independent of the speed ring power regulation logic. This double-channel decoupling control strategy maintains the constant cutting power by varying the speed, and uses virtual stiffness injection to offset the mechanical tool compensation error caused by cutting force fluctuation, so that the tooth profile geometric accuracy can be maintained when machining high-hardness or large-modulus hanging wheels on rigid-limited machine tools. BRIEF DESCRIPTION OF DRAWINGS

[0021] Fig. 1 The system control principle block diagram of the angle domain mapping and phase lead constraint of the present application.

[0022] Fig. 2 The static angle load topology curve of the present application containing the distribution of machining features.

[0023] Fig. 3The system hardware level architecture and zero-time difference closed loop data interaction topology of the application. DETAILED DESCRIPTION

[0024] The technical solutions in the embodiments of the application will be clearly and completely described below with reference to the embodiments of the application. Obviously, the described embodiments are only part of the embodiments of the application, rather than all the embodiments of the application. Based on the embodiments of the application, all other embodiments obtained by those skilled in the art without creative work fall within the protection scope of the application.

[0025] The hanging wheel assembly processing process parameter intelligent optimization system provided by the application is composed of four core function modules, i.e., a multi-dimensional data synchronous acquisition unit, an angle domain static atlas construction unit, a phase lead constraint control unit and an angle trigger execution unit. The modules perform data interaction through a high-speed field bus or a shared memory mechanism, forming a closed loop feedforward control topology with the spindle angle as an absolute reference and the load current as a controlled object. When controlling the high-frequency load impact in the hanging wheel processing process, due to the inherent physical time constant of the current loop filtering of the servo motor, the mechanical transmission chain and the numerical control system, the traditional control strategy based on real-time feedback cannot complete the response in the millisecond-level intermittent cutting-in instant. The system adopts a periodic load feedforward compensation mechanism based on angle domain mapping. The multi-dimensional data synchronous acquisition unit accesses the current monitoring port of the machine tool servo driver and the high-speed pulse output port of the spindle position encoder. The unit is configured with a hardware parallel latching logic based on a field programmable gate array (FPGA). The hardware interrupt signal is used to synchronously trigger the acquisition of the spindle load current time sequence and the spindle angle position sequence, so as to ensure that the time deviation between the load current sampling time and the position encoder counting time is strictly controlled within one tenth of the servo control period, thereby providing a high-fidelity synchronous data source for subsequent signal reconstruction.

[0026] The angle domain static atlas construction unit performs dimension transformation processing from the time domain to the angle domain. For the non-periodic random noise existing in the original acquisition data, the unit adopts a periodic superposition synchronous average algorithm. Specifically, the system obtains original cutting data streams of multiple spindle rotation periods. The load data at the same angle position is accumulated and the arithmetic mean value is taken with the spindle angle as the index. Through this processing, the random disturbance in the cutting process is eliminated, and a static angle-load topology atlas capable of accurately representing the load change law of the tool cutting-in point, cutting-out point and hard point position is generated. The atlas takes the spindle angle as the only independent variable index, and takes the load current characteristic value as the dependent variable index. For mapping objects, so as to eliminate time dimension variable in control logic, establish the deterministic space-load mapping relationship; phase advance constraint control unit is responsible for generating the control instruction with the predictive, for the physical delay existing between the servo system from receiving instruction to output torque, the unit has stored in advance the system inherent total lag time including current loop filter delay time, operation interpolation period and mechanical transmission chain inertia hysteresis constant In the control cycle, the unit reads the current angular velocity of the main shaft in real time And according to the formula Real-time solution phase advance angle The unit identifies the load mutation characteristic area according to the static angle-load topological atlas, and generates the corresponding parameter adjustment instruction such as feed ratio adjustment or torque compensation instruction for the area, the unit shifts the execution trigger time of the above instruction forward by phase advance angle in the angle domain coordinate system Thus, the parameter adjustment instruction sequence with negative time axis offset attribute is generated, which ensures that the control action has started before the time when the physical load impact occurs, and offsets the response lag of the physical system.

[0027] For the generation of feed ratio adjustment instruction in parameter adjustment instruction sequence, the phase advance constraint control unit is configured with inverse mapping logic based on constant power cutting principle, aiming to establish a nonlinear inverse proportional relationship between feed ratio and load characteristic value in static angle-load topological atlas, the system sets the reference cutting load target value And the maximum allowable feed ratio For each discrete angle index In the atlas , the corresponding feed ratio adjustment instruction is calculated, which follows the inverse proportional gain function Wherein Is the load current characteristic value of the atlas at this angle, The power balance coefficient for adjusting response sensitivity, when the atlas shows that there is a high load peak in a certain angle interval, the function automatically generates a feed ratio instruction lower than To limit the cutting force; when in low load interval, generate higher feed ratio instruction to improve machining efficiency, the generated ratio instruction sequence is phase shifted to The moment is the start of the deceleration action, so as to maintain the dynamic constant of the instantaneous cutting power; the angle trigger execution unit acts as a real-time scheduler during the machining operation, which locks the current angular position of the main shaft in real time and compares it with the instruction sequence index generated by the phase lead constraint control unit. When the main shaft rotates to the preset angular window before the physical load impact occurs, that is, the actual angle reaches the target feature angle minus the phase lead angle position, the unit directly bypasses the conventional speed loop feedback logic, injects the parameter adjustment instruction sequence into the current loop sum point of the servo system, and through this procedure, the servo motor has completed the establishment of the pre-acceleration or pre-deceleration electromagnetic torque at the moment when the tool actually contacts the workpiece gear slot, realizing zero-time difference torque compensation for the periodic intermittent cutting load.

[0028] In view of the individual differences of the workpiece caused by the installation eccentricity of the hanger wheel blank or the batch fluctuation of the material hardness, the angular domain static atlas construction unit is further configured with atlas dynamic migration logic. In the first cutting period of the machining process of each new workpiece, the logic starts the feature anchor point capture module, which monitors the first derivative of the load current to the angle in real time and locks the moment when the first rising edge of the derivative exceeds the preset trigger threshold, records the actual instantaneous phase angle and the actual peak load at that moment, the deviation operator solution module calls the corresponding reference cutting phase angle and the reference peak load in the static angular-load topological atlas, calculates the phase offset and the amplitude proportional coefficient , the execution atlas correction module performs a translation transformation on the angle index axis of the original atlas according to the principle of affine transformation, and performs a linear gain scaling on the load value in the atlas using the amplitude scale factor, to generate a temporary execution atlas suitable for the current specific workpiece. This procedure does not require full-cycle relearning for each workpiece to achieve transient adaptation of the control model to individual differences; To avoid the problem of current sensor zero drift caused by temperature changes in the workshop environment, the angular domain static atlas construction unit has a dynamic zeroing logic for the idle period window. The non-cutting idle period that exists during the intermittent cutting process is used to identify the continuous angle interval in the static angle-load topology atlas where the load current and its rate of change are both below the preset idle threshold, and this interval is marked as the zero point calibration window. In each subsequent spindle rotation period, when the spindle angle enters the zero point calibration window, the dynamic bias update module triggers high-frequency sampling and calculates the arithmetic mean of the current data in the window, which is used as the current direct current component drift value. The real-time correction module subtracts the most recently updated direct current component drift value from the real-time collected load current signal when the spindle rotation angle enters the cutting work interval. This dynamic calibration mechanism eliminates the influence of thermal drift on the cutting load measurement accuracy without the need for an external constant temperature reference source, ensuring the consistency of the control system during the cold start and hot state balancing processes.

[0029] To solve the problem of elastic deformation of the process system caused by cutting force fluctuation, which affects the profile accuracy, a virtual stiffness compensation unit is integrated into the system. The unit pre-stores the equivalent stiffness coefficients of the controlled object obtained through static stiffness experiment calibration During the machining process, the deformation estimation module calculates the expected elastic deformation at the current angle according to the load current value in the static angle-load topology atlas using the inverse model of Hooke's law wherein is the current load current value The superimposed control injection module generates a micro-displacement compensation instruction with the same amplitude but opposite direction according to the expected elastic deformation , which is superimposed to the position loop control node of the feed axis servo driver. The direct intervention of this position loop is independent of the power regulation logic of the speed loop. Through the injection of micro-scale reverse displacement, the mechanical tool deflection error caused by physical cutting force is offset, maintaining the geometric position accuracy of the system; To solve the problem of nonlinear increase of cutting resistance caused by tool wear, which leads to insufficient feedforward compensation, the phase lead constraint control unit is configured with a gain adaptive evolution module. The residual extraction submodule in this module calculates the difference between the measured load current and the preset target load value in each machining period to generate a control residual sequence. The energy integration submodule selects the control residual sequence within the cutting angle interval to calculate the integral of the absolute value and generate a periodic residual energy index , which quantifies the degree of mismatch between the current feedforward model gain and the actual cutting state, the gain scheduling submodule updates the global feedforward gain coefficient using a cumulative step algorithm, and the update rule follows the formula , where is the updated gain coefficient for the next period, is the current period coefficient, is a preset learning rate step size, and the phase-advance constraint control unit dynamically scales the amplitude of the parameter adjustment instruction sequence using the real-time updated global feedforward gain coefficient, thereby automatically compensating for the cutting resistance increment caused by tool wear and maintaining constant control effect throughout the life cycle.

[0030] Embodiment 1: In the application scenario of high-speed intermittent cutting of large modulus hanging wheel components, the workpiece material is high-strength alloy steel, the machining process is accompanied by periodic severe cutting force impact, and the spindle speed is set in the interval of 200 to 500 rpm, resulting in a time interval of adjacent tooth groove cutting action shortened to milliseconds. Under this working condition, the conventional servo control system is limited by current loop filter delay and mechanical transmission chain inertia, and cannot output compensation torque in time at the moment of tool cutting, resulting in a transient drop in spindle speed and causing machining vibration marks. This embodiment uses a multi-dimensional data synchronous acquisition unit to access the servo driver and position encoder. In the first trial cutting stage, the hardware parallel latching logic of the field programmable gate array (FPGA) is used to synchronously acquire the spindle load current time series and the spindle angle position sequence with a high time resolution of one tenth of the servo control period. The angle domain static atlas construction unit then performs time domain to angle domain transformation on the synchronous data, processes data of 5 to 10 consecutive rotation periods using a periodic superposition synchronous averaging algorithm, filters out random noise, and generates a unique index representing the static angle-load topological atlas of the specific hanging wheel cutting force variation law , where is the real-time angular velocity of the spindle, and the system inherent total lag time previously calibrated and stored is called, and the phase-advance angle is calculated in real time according to the formula . The system identifies the cutting-in feature area of the load current steep rise from the static angle-load topological atlas, generates the corresponding torque compensation instruction, and translates the trigger position of the instruction in the angle coordinate system by , angle trigger execution unit locks the spindle position in real time, and directly injects the torque compensation instruction into the servo current loop within a preset angle window before the occurrence of physical cutting impact. This feedforward control mechanism enables the servo motor to establish an electromagnetic torque matching the cutting resistance at the physical moment when the tool actually contacts the workpiece, controls the spindle speed fluctuation amplitude within 0.1% of the rated speed, and avoids tooth profile errors caused by response lag.

[0031] Example 2: To verify the effectiveness of the technical scheme of the present application in actual engineering environment, a test verification platform containing real physical load and environmental interference is built. The platform takes a high-speed gear hobbing machine equipped with a Fanuc 0i-MF numerical control system as the core, the spindle drive motor has a rated power of 15 kW and a maximum speed of 6000 rpm, the experimental workpiece is a 20CrMnTi alloy steel hanging wheel blank with a module of 4 and a tooth number of 48, to simulate the electromagnetic interference in real industrial field, a power frequency interference signal with a frequency of 50 Hz and an amplitude of 0.5 A is actively superimposed on the current feedback loop of the servo driver, and a Gaussian white noise with a signal-to-noise ratio of 20 dB is introduced, and to verify the stability of the system to environmental thermal drift, the temperature in the experimental workshop is periodically fluctuated between 15 to 35 within 24 hours through the environmental control system.

[0032] The test process is as follows: in the comparison sample group, the multi-dimensional data synchronous acquisition unit and the phase advance constraint control unit of the present application are turned off, and only the speed loop and current loop feedback logic built in the numerical control system is used for cutting, in the sample group of the present application, the whole set of intelligent optimization system is enabled, the system performs data acquisition of no-load and trial cutting in the first rotation period, constructs the angle domain static atlas, the phase advance constraint control unit calculates the phase advance angle according to the real-time speed and the calibrated total lag time , and generates the feedforward torque instruction; to quantitatively evaluate the control effect, the angular velocity fluctuation of the spindle is monitored in real time by a high-precision laser vibration meter, and the actual q-axis current response of the servo motor is recorded by a current probe, the comparison and analysis of the key intermediate data reveal the internal working mechanism of the present application, as shown in Table 1, in the comparison sample group, at the moment when the tool cuts into the workpiece (time point ), due to the physical response lag, the establishment of the electromagnetic torque lags behind the load mutation by about 12 ms, causing the instantaneous drop of the spindle speed, and the peak-to-peak value of the speed fluctuation reaches 18.5 rpm, while in the sample group of the present application, thanks to the introduction of the phase advance angle, the electromagnetic torque begins to establish at the moment before the tool cuts in, and reaches the load balance point at the moment of cutting in ( ).

[0033] Table 1: Comparison of key performance indicators under different control strategies

[0034] Furthermore, to verify the effectiveness of the dynamic migration logic of the graph and the adaptive evolution module of the gain, a long-cycle continuous machining test was conducted. When machining the 50th workpiece, a workpiece eccentricity error of 0.05 mm was artificially introduced. The rotational speed fluctuation of the comparison sample group deteriorated to more than 25 rpm, while the sample group of the present invention, through feature anchor point capture and graph correction, brought the rotational speed fluctuation back to within 0.5 rpm within 2 cutting cycles. As tool wear increased (to the 150th piece), the machining error of the comparison sample group showed a linear divergence trend, while the sample group of the present invention used residual energy integral feedback to dynamically increase the feedforward gain coefficient, so that the tooth profile error of the 150th piece was still kept within the accuracy range of 4.0 μm, verifying the adaptive capability of the system throughout its entire life cycle.

[0035] Example 3: This example combines Figs. 1 to 3 The explanation of the intelligent optimization system for the processing parameters of the train roller assembly is as follows: Fig. 1 As shown, the system architecture demonstrates a complete closed-loop logic flow from signal acquisition to command execution. The starting end consists of a servo driver and a position encoder, responsible for outputting the original current sequence, spindle angular position, and real-time angular velocity signals. These signals are input to a multi-dimensional data synchronous acquisition unit, which transmits the data to the angular domain static map construction unit through an FPGA hardware parallel latch and hard synchronization mapping interface. Here, the transformation from the time domain to the angular domain and the generation of the static angle-load topology are completed. The data flows to the phase lead constraint control unit, which is responsible for calculating the phase lead angle and generating a feedforward command sequence. Finally, the angle trigger execution unit injects commands into the preset angle window to achieve zero-time-difference torque compensation. At the same time, the system is supplemented by map dynamic migration logic to handle individual differences in the workpiece and feature anchor point capture, configured idle window dynamic zeroing logic to eliminate environmental thermal drift, set up a gain adaptive evolution module to compensate for tool wear and perform residual energy integration, and integrate a virtual stiffness compensation unit to offset elastic deformation and inject micro-displacement compensation. All modules work together to achieve feed rate adjustment and physical load impact compensation at the servo current loop control node.

[0036] like Fig. 2 As shown, this graph is built in a two-dimensional coordinate system. The horizontal axis represents the spindle angle, covering a complete rotation cycle from 0° to 360°. The vertical axis represents the load current, with a scale range from 0A to 20A. The core of the graph is a static angle trajectory curve formed by a dashed line, vividly illustrating the load variation during the gear cutting process. The entry feature zone near 80° is clearly marked, where the load current shows an upward trend. It reaches a hard point at 170°, corresponding to the peak load current region. The curve enters the exit feature zone at 260°, where the load current decreases accordingly. Fig. 3As shown, the entire system is divided into three core parts: the physical processing and driving end, the high-speed hardware synchronization end, and the intelligent optimization calculation center. The physical processing and driving end includes servo motors, encoders, and intelligent servo drivers. Its output position and current signal sequences are transmitted to the high-speed hardware synchronization end through the dotted line. This end is equipped with an FPGA parallel latch terminal to process the hard synchronization mapping data stream. The data is sent to the intelligent optimization calculation center, i.e., the industrial control host, which includes an angle domain static map construction engine and a phase advance constraint control core, for processing. The calculated feedforward torque compensation command with preset angle window injection attributes is fed back to the physical processing and driving end, thereby constructing a zero-time-difference closed-loop control loop.

[0037] Example 4: Addressing the problem that traditional PI control parameter tuning relies on manual experience and is difficult to adapt to time-varying operating conditions during gear machining, this example constructs a gain adaptive evolution procedure based on residual energy index. Instead of relying on complex mathematical models for identification, it utilizes the error signals naturally generated during system operation to achieve online self-optimization of the feedforward gain. In the initial stage of system operation, a parameter initialization step is performed, and the system adjusts the global feedforward gain coefficients... The initial value is set to and adjust the gain correction step size Set as The initial value is chosen based on the theoretical stiffness model of the servo system, while the step size is set to balance convergence speed and stability, avoiding system oscillations caused by excessively large step sizes. The system enters a periodic iterative optimization process, and at the 1st... Within each processing cycle, the phase lead constraint control unit adjusts its control based on the current gain coefficient. Generate feedforward commands and collect the actual load current in real time during the cutting process. With target load current The system calculates the difference between the two, i.e., the control residual. To quantitatively assess the matching degree of the current gain coefficient, the system calculates the cutting angle range for that cycle. Residual energy index within The calculation formula is: This indicator directly reflects the total deviation between the feedforward compensation and the actual cutting resistance. A large value indicates that the current feedforward gain is insufficient to offset the cutting force or is too large, leading to overcompensation.

[0038] Based on the calculated residual energy index, the gain scheduling submodule updates the gain coefficient according to the gradient descent principle, with the update rule being: ,in The sign function of the average residual for this period is given. If the average residual is positive (i.e., the actual load is greater than the target load), it indicates insufficient compensation, and the gain coefficient will increase; conversely, it will decrease. The system continues to execute the above iterative process until continuous... a rate of change of the residual energy index over a period of time is less than a preset convergence threshold, such as 2%, at which point it is determined that the gain coefficient has converged to an optimal value for the current operating condition, when a change in operating condition is detected, such as a change in tool or material batch mutation, the system automatically resets the step size and restarts the fast search process.

[0039] Example 5: Before deploying the intelligent optimization system for machining process parameters of the wheelset assembly to an actual production line, a standardized system initialization calibration and parameter calibration procedure is performed to ensure accurate matching of the control strategy with the physical characteristics of the specific machine tool, which aims to quantitatively determine key physical parameters in the control model, including the total lag time of the servo system and the equivalent stiffness coefficient of the process system Under the no-load state of the machine tool, the spindle is rotated at a series of step speeds, while a sweep current signal with a frequency of 10 Hz to 1000 Hz is injected into the servo driver. The time difference between the current command and the actual current response is recorded using a multi-dimensional data synchronous acquisition unit. By weightedly averaging the phase lags at different frequency points, the total inherent lag time of the system, including the current loop filter delay and the inertia of the mechanical transmission chain, is determined .

[0040] Static calibration of the stiffness of the process system is performed. Under the stationary state of the spindle, a series of known static radial forces are applied to the spindle using a force sensor, and the elastic deformation of the spindle in the force direction is measured using a high-precision displacement sensor. By recording multiple sets of force-displacement data points, a linear regression curve is fitted using the least squares method, and the slope is the equivalent stiffness coefficient of the process system This coefficient is stored in the virtual stiffness compensation unit and used to convert the load current value into the expected elastic deformation in real time during cutting. In addition, before machining a specific type of wheelset for the first time, the system performs a free travel feature learning process. The spindle rotates slowly under the no-load state, and the angular domain static atlas construction unit records the no-load current reference curve during this process, identifies the inherent load fluctuation features caused by mechanical assembly errors or uneven friction, and deducts them in subsequent cutting load monitoring.

[0041] Example 6: During the engineering configuration stage before the system processes a batch of specific types of wheelsets, a standardized identification procedure including current-force conversion coefficient calibration and angle resolution definition is performed to eliminate the control model uncertainty introduced by differences in machine tool motor characteristics and workpiece geometric features. The current-force conversion coefficient in the virtual stiffness compensation unit , the system guides the external standard dynamometer to apply a stepwise increasing static load covering the range of the rated cutting force in the tangential direction of the workpiece's index circle under the main shaft servo locking state , synchronously latches the q-axis current value fed back by the servo driver , and performs a least square linear regression operation on the obtained multiple sets of load-current data pairs, and solidifies the slope of the obtained regression straight line as the dedicated conversion coefficient of the machine tool , thereby establishing a deterministic mapping reference between the current signal and the physical cutting force.

[0042] Meanwhile, to ensure that the graph generated by the angular domain static graph construction unit can restore the microscopic load characteristics in the intermittent cutting process without distortion, the system determines the feature fidelity constraint according to the number of teeth of the workpiece to be machined and the characteristic fidelity constraint determined by the Shannon sampling theorem, calculates the optimal angular discretization resolution using the formula , where , and is the minimum number of sampling points required to analyze the load waveform within a single pitch and is not less than 20, and the system divides the angular index address in the storage space of the controller according to the calculation result to establish a high-precision data container adapted to the geometric characteristics of the current workpiece.

[0043] It is obvious to those skilled in the art that the present application is not limited to the details of the above exemplary embodiments, and can be implemented in other specific forms without departing from the spirit or essential characteristics of the present application.

[0044] Finally, it should be noted that the above examples are only used to illustrate the technical solutions of the present application and are not limiting, and although the present application has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present application can be modified or replaced by equivalents without departing from the spirit and scope of the present application.

Claims

1. A hanging wheel assembly processing process parameter intelligent optimization system, characterized in that, The application relates to a servo control system for a workpiece cutting machine, which comprises the following units: a multi-dimensional data synchronous acquisition unit, which is connected to a servo driver and a position encoder, uses a hardware interrupt signal to trigger parallel latching logic, and establishes a hard-synchronous mapping interface between a spindle load current time sequence and a spindle angle position sequence; an angle-domain static atlas construction unit, which performs dimension transformation from a time domain to an angle domain, generates a static angle-load topology atlas with a spindle angle as a unique index and a load current characteristic value as a mapping object through phase alignment and statistical filtering of multi-cycle synchronous data, and uses the load topology atlas to eliminate time dimension variables of servo control; a phase lead constraint control unit, which calculates a phase lead angle based on a product of inherent physical response lag time of the servo system and real-time angular velocity of the spindle, and generates a parameter adjustment instruction sequence with a negative time axis offset attribute according to a distribution position of a load mutation feature area in the static angle-load topology atlas; and an angle trigger execution unit, which locks a current angle position of the spindle in real time during control operation, is controlled by the phase lead angle, directly injects the parameter adjustment instruction sequence into a servo current loop in a preset angle window before a physical load impact occurs, and realizes zero-time-difference torque compensation for a periodic intermittent cutting load at a physical level. The angle-domain static atlas construction unit further comprises atlas dynamic migration logic for processing workpiece individual differences, which comprises a feature anchor point capture module, a deviation operator calculation module, and an execution atlas correction module. The feature anchor point capture module monitors a load current first-order derivative in a first waveform of a control cycle, locks a transient phase angle of an actual waveform rising edge and an actual peak load; the deviation operator calculation module calculates a phase offset between the transient phase angle and a reference phase angle in the static angle-load topology atlas, and calculates an amplitude proportional coefficient between the actual peak load and a reference peak load; and the execution atlas correction module performs a translation transformation on an index axis of the static angle-load topology atlas by using the phase offset, and performs a linear gain scaling on atlas load values by using the amplitude proportional coefficient, to generate a temporary execution atlas adapted to a current control object.

2. The system of claim 1, wherein, The angle-domain static atlas construction unit further comprises an idle stroke window dynamic zeroing logic for eliminating environmental thermal drift interference, which comprises a window locking module, a dynamic bias updating module, and a real-time correction module. The window locking module identifies a non-working angle interval with a load current and a change rate of the load current both lower than a preset threshold in the static angle-load topology atlas, and marks the interval as a zero-point calibration window; the dynamic bias updating module collects real-time load current and calculates a direct current component drift value when the spindle rotates to the zero-point calibration window; and the real-time correction module subtracts the direct current component drift value calculated last time from the real-time collected load current signal when the spindle rotates to a working angle interval, and outputs corrected load feedback data.

3. The system of claim 1, wherein, ​ 4. The system of claim 1, wherein, The system further comprises a virtual stiffness compensation unit for offsetting the mechanical elastic deformation error, the unit comprising: a stiffness model storage module storing equivalent stiffness coefficients of the controlled object; a deformation amount estimation module for calculating the expected elastic deformation amount of the controlled object at the current angle based on the load current value in the static angle-load topological atlas and the equivalent stiffness coefficients; and a superimposed control injection module for generating a micro-displacement compensation instruction with the same magnitude and opposite direction of the expected elastic deformation amount, and directly superimposing the instruction to the position loop control node of the feed shaft, independent of the speed loop control instruction to maintain the geometric position accuracy of the system.

5. The system of claim 1, wherein, The phase lead constraint control unit further comprises a gain adaptive evolution module for compensating for the resistance change caused by tool wear, the module comprising: a residual extraction submodule for calculating the control residual sequence between the measured load current and the preset target load value in each machining period; an energy integration submodule for performing integral operation on the control residual sequence in the cutting angle interval to generate a period residual energy index representing the mismatch degree of the feedforward model; and a gain scheduling submodule for updating the global feedforward gain coefficient based on the period residual energy index, and the phase lead constraint control unit uses the coefficient to dynamically scale the amplitude of the parameter adjustment instruction sequence.

6. The system of claim 5, wherein, The gain scheduling sub-module adopts a cumulative step algorithm based on residual energy to update the global feedforward gain coefficient, and the update rule is: wherein, is the updated global feedforward gain coefficient, is the global feedforward gain coefficient of the current period, is a preset learning rate step, is a period residual energy index, which is obtained by integrating the absolute value of the control residual sequence in the cutting interval.

7. The system of claim 1, wherein, The phase lead angle in the phase lead constraint control unit is determined based on the product of the sum of the current loop filter delay time of the servo system, the interpolation calculation period and the inertia lag constant of the mechanical transmission chain and the real-time angular velocity of the main shaft, and the phase lead angle defines the advance amount of the control instruction relative to the position of the physical load in the angle domain coordinate system.

8. The system of claim 1, wherein, The multi-dimensional data synchronous acquisition unit uses a field programmable gate array (FPGA) as a hardware carrier, and triggers parallel latching of data through a hardware interrupt signal.

9. The system of claim 1, wherein, The parameter adjustment instruction sequence includes a feed rate adjustment instruction, and the phase lead constraint control unit is configured with inverse mapping logic, which establishes an inverse proportional gain function between the feed rate and the cutting load, generates a low feed rate instruction in the high load characteristic region of the static angle-load topological atlas, and generates a high feed rate instruction in the low load characteristic region.

10. The system of claim 1, wherein, The angle domain static atlas construction unit comprises a period superposition submodule for obtaining original cutting data streams of a plurality of continuous rotation periods, and performing superimposed average processing on load data at the same angle index position by using a synchronous average algorithm.

Images