A large numerical control cutting saw cutting precision intelligent regulation method and system

Abstract

The present application belongs to the technical field of precision sawing machine, and particularly relates to a large-scale numerical control cutting sawing machine cutting precision intelligent regulation and control method and system, which comprises the following steps: calculating resistance mutation and inertia response characteristic items based on current fluctuation and acceleration characteristics, introducing a zero micro value prevention to obtain sawing dynamic hysteresis time by combining sawing band suspension length and feed speed; calculating space-time dislocation compensation span by combining system response time delay, feed speed and sawing band torsional deformation angle, and then extracting feed correction gain coefficient through band limiting constraint nonlinear mapping; inputting multi-modal data into a deep learning model to obtain a basic compensation amount, and using the gain coefficient to dampen the feed speed and amplify the tension force, so as to complete the compensation correction of the feed speed and the tension force. The present application avoids mechanical response lag and space-time dislocation caused by cutting resistance mutation, suppresses compensation overload under extreme working conditions, and improves heavy load cutting precision and production safety.

Description

Technical Field

[0001] This invention relates to the field of precision sawing technology. More specifically, this invention relates to an intelligent control method and system for the cutting precision of a large CNC cutting saw. Background Technology

[0002] Large CNC sawing machines are key basic equipment in the metal processing field used for cutting and shaping. They mainly achieve the cutting of various bars, tubes, and irregularly shaped metal materials through the high-speed rotation of the circular saw blade and the propulsion of the hydraulic feed system. Real-time monitoring and adjustment of the sawing machine's feed speed and saw blade tension by the CNC system is the core means to ensure cutting accuracy and stable operation of the equipment.

[0003] In actual heavy industrial production scenarios, these large sawing machines are often used for high-speed, heavy-duty machining of large or complex metal forgings. When cutting these large metal parts, due to the unavoidable uneven hardness distribution or abrupt changes in cross-sectional width within the material, the saw blade frequently encounters severe transient changes in resistance during the cutting process. Simultaneously, hydraulic servo actuators in industrial settings generally suffer from mechanical inertia caused by the mass of the transmission chain, and the inherent response delay of the system always exists between the issuance of control commands and their physical execution.

[0004] During the heavy-duty cutting of large metal parts, when encountering sudden changes in material cutting resistance, the traditional control method cannot predict and dynamically compensate for the unavoidable mechanical response lag and time gap between the issuance of CNC commands and the execution of underlying hydraulic hardware. This causes the saw blade to continue to advance within the metal at the original feed rate during this lag period, resulting in severe nonlinear skew and torsional deformation of the saw blade and spatiotemporal misalignment in the cutting trajectory. This not only directly leads to the scrapping of high-value large metal workpieces due to out-of-tolerance cutting sections, but also easily causes serious production safety accidents such as saw blade jamming and breakage. Summary of the Invention

[0005] To address the aforementioned technical problem of poor cutting control effect in large CNC cutting saws, this invention provides solutions in the following aspects.

[0006] In a first aspect, the present invention provides a method for intelligent control of cutting accuracy of a large CNC cutting saw, comprising: The system acquires the current and reference current of the spindle motor of the sawing machine at various moments, the feed speed and acceleration of the sawing machine, the skew displacement and torsional deformation angle of the saw blade; it obtains the inherent system response delay of the sawing machine through system calibration, the saw blade overhang length, the material thickness of the workpiece being cut, the maximum safe skew displacement threshold of the sawing machine, and the sampling time interval of the CNC system and the machine tool reference speed of the sawing machine; based on the current difference of the spindle motor at adjacent moments and the feed acceleration of the sawing machine, combined with the saw blade overhang length and the feed speed of the sawing machine, it obtains the sawing dynamic hysteresis time of the CNC sawing machine; and it combines the sawing dynamic hysteresis time with the inherent system response delay. The model is magnified and projected using the feed speed and torsional deformation angle of the saw to obtain the spatiotemporal misalignment compensation span of the saw. Based on the offset ratio of the spatiotemporal misalignment compensation span relative to the material thickness of the currently cut workpiece, combined with the skew displacement of the saw blade, the feedforward correction gain coefficient of the CNC system is obtained. The current of the spindle motor, the feed speed of the saw, the feed acceleration of the saw, the skew displacement of the saw blade, and the torsional deformation angle of the saw blade are input into the deep learning model to output the basic feed speed compensation amount and the basic tension compensation amount of the saw. The model is then adjusted using the feedforward correction gain coefficient to obtain the final feed speed compensation amount and the final tension compensation amount of the saw.

[0007] This invention, by fusing multi-source sensor data, dynamically calculates the underlying mechanical hysteresis characteristics and spatiotemporal misalignment compensation span, accurately assessing this physical hysteresis gap. This avoids blind feeding of the machine tool when facing sudden resistance, reducing the risk of transient overload on the spindle motor and hydraulic system. Combining the inference capabilities of a temporal deep learning model with a feedforward correction mechanism, this invention transforms the calculated spatiotemporal misalignment span and skew state into feedforward gain coefficients. This enables the sawing control system to anticipate and proactively reduce the feed speed and simultaneously amplify the saw band tension before encountering extreme cutting resistance that leads to deformation deterioration, achieving dynamic adaptive flexible control.

[0008] Preferably, obtaining the dynamic hysteresis time of the CNC saw includes: Obtain the resistance mutation characteristic of the sawing process; obtain the inertial response characteristic of the feed system; multiply the ratio of the resistance mutation characteristic at any time to the inertial response characteristic at that time by the product of the suspended length of the saw band and the feed speed of the sawing machine at that time to obtain the sawing dynamic hysteresis time of the CNC sawing machine at that time.

[0009] Preferably, the spatiotemporal misalignment compensation span satisfies the expression: ; In the formula, This represents the span of the spatiotemporal misalignment compensation for the saw at time t. This represents the feed rate of the saw at time t. This represents the inherent system response delay of the saw; This represents the dynamic hysteresis time of the CNC saw at time t. This represents the torsional deformation angle of the saw blade at time t. This represents the torsional attenuation projection constant.

[0010] This invention superimposes the dynamic hysteresis phenomenon into the inherent response delay of the system, and combines the actual torsional posture of the saw belt to perform an effective algebraic attenuation projection of the physical sliding distance. In high-speed propulsion cutting scenarios where there is a control gap in the hydraulic hardware, it restores the real spatial physical offset caused by mechanical hysteresis and posture deflection, ensuring the accuracy of subsequent control commands in alignment correction in both spatial and temporal dimensions.

[0011] Preferably, the resistance mutation characterization term satisfies the expression: ; In the formula, The term representing the sudden change in resistance at time t during the sawing process; , This represents the current of the spindle motor at time t and time t-1; This indicates the reference current of the spindle motor; Represents the absolute value function; This represents the resistance change sensitivity constant.

[0012] This invention employs a purely linear dimensionless calculation method that combines the division of reference currents. In the heavy forging cutting environment where the spindle motor faces high-frequency transient load impacts, it not only filters out electrical background noise under normal machining conditions, but also robustly and smoothly assesses the sharp increase in cutting resistance, preventing sudden data changes from causing direct failure of the algorithm control system.

[0013] Preferably, the inertial response characterization term of the feed system satisfies the expression: ; In the formula, This represents the inertial response characteristic of the feed system at time t. This represents the feed acceleration of the saw at time t. Indicates the sampling time interval of the CNC system; This indicates the machine tool reference speed of the saw; This represents the preset inertial response sensitivity constant.

[0014] This invention combines microscopic feed acceleration with sampling time interval and performs dimensionless processing using machine tool reference speed. It directly maps physical response characteristics using pure algebraic addition and multiplication. Under different harsh working conditions and huge transmission chain inertia scenarios faced by CNC sawing machines, it characterizes the objective resistance inertia level exhibited by the feed system when facing sudden changes in resistance, thereby improving the robustness of the underlying physical feature modeling.

[0015] Preferably, the feedforward correction gain coefficient of the CNC system satisfies the expression: ; In the formula, This represents the feedforward correction gain coefficient of the CNC system at time t; This represents the span of the spatiotemporal misalignment compensation for the saw at time t. Indicates the material thickness of the workpiece currently being cut; The absolute value of the skew displacement of the saw blade at time t is represented by a scalar. This indicates the maximum safe skew displacement threshold for the saw.

[0016] This invention utilizes the actual thickness of the processed material to perform an inverse proportional saturated nonlinear scale transformation, and introduces a pure linear proportional operator based on the maximum safety threshold constraint. In critical cutting situations where the saw blade is laterally deflected or on the verge of breaking, it can adaptively output a smooth and reasonably strong amplification multiplier, avoiding the mechanical oscillation and secondary damage caused by overcompensation under traditional linear control.

[0017] Preferably, the adjustment using the feedforward correction gain coefficient includes: The ratio of the basic feed rate compensation of the saw to the feedforward correction gain coefficient is used as the final feed rate compensation of the saw. The product of the basic tension compensation amount and the feedforward correction gain coefficient is used as the final tension compensation amount of the saw belt.

[0018] Preferably, the acquisition of the current and reference current of the saw spindle motor at various times, the feed speed of the saw, the feed acceleration of the saw, the skew displacement of the saw band, and the torsional deformation angle includes: The current of the main spindle motor is collected by a current sensor installed on the main spindle motor of the saw; the reference current of the main spindle motor is obtained by the no-load operation test of the saw; the feed speed of the saw is collected by an optical encoder installed on the feed drive shaft of the saw; the feed acceleration of the saw at each moment is obtained based on the ratio of the difference in feed speed at adjacent moments to the sampling time interval; the skew displacement of the saw band at each moment is obtained by a displacement sensor installed on the guide arm of the saw, and the torsional deformation angle of the saw band at each moment is obtained by differential calculation of dual-channel sensor data.

[0019] Preferably, the acquisition of the inherent system response delay of the sawing machine, the acquisition of the saw blade overhang length, the material thickness of the current workpiece being cut, the maximum safe skew displacement threshold of the sawing machine, and the sampling time interval of the CNC system and the machine tool reference speed of the sawing machine includes: The inherent response delay of the system is obtained through the system calibration module of the saw; the overhead length of the saw band, the sampling time interval of the CNC system, and the machine tool reference speed of the saw are obtained through the machine tool parameter reading module; the material thickness of the workpiece being cut is obtained through the machine tool parameter setting interface; and the maximum safe deviation displacement threshold of the saw set by the CNC system is obtained.

[0020] Secondly, the present invention provides an intelligent control system for the cutting precision of a large CNC cutting saw, comprising a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the above-mentioned intelligent control method for the cutting precision of a large CNC cutting saw is realized.

[0021] By adopting the above technical solution, a computer program is generated from the above-mentioned intelligent control method for cutting precision of a large CNC cutting saw, and stored in the memory so that it can be loaded and executed by the processor. In this way, a terminal device can be made based on the memory and the processor for convenient use.

[0022] The beneficial effects of this invention are as follows: (1) The control mechanism of the present invention provides the saw blade with sufficient elastic reset space and rigid support tension when the cutting resistance changes suddenly by using a dual-track opposite adjustment strategy of feed deceleration and tension burst pull. This suppresses the nonlinear lateral deviation and longitudinal torsional deformation of the saw blade inside the heavy metal, ensuring the flatness and perpendicularity of the cutting section and avoiding the scrapping of high-value large metal workpieces due to cutting deviation. (2) The present invention can monitor the torsional deviation of the saw blade in real time and combine it with the maximum safe deviation displacement threshold to prevent overload constraint and limit it. This avoids mechanical vibration caused by overcompensation and prevents the saw blade from getting stuck, breaking the saw blade and chipping teeth under extreme working conditions. This not only ensures the safety of human and machine operation in heavy industrial production sites, but also extends the service life of high-speed steel / carbide saw blades and core transmission components of machine tools, and reduces the overall operation and downtime maintenance costs of enterprises. Attached Figure Description

[0023] Figure 1 This is a flowchart illustrating an intelligent control method for cutting precision of a large CNC cutting saw in this invention; Figure 2 This is a schematic diagram showing the change of dynamic hysteresis time in sawing over time; Figure 3 This is a schematic diagram showing the change of the feedforward correction gain coefficient of the CNC system over time. Detailed Implementation

[0024] This invention discloses an intelligent control method for cutting accuracy of a large CNC cutting saw, referring to... Figure 1 This includes steps S1-S4: S1: Obtain multi-dimensional data for the cutting control of the sawing machine. The specific steps are as follows: obtain the current, feed speed, feed acceleration, skew displacement, and torsional deformation angle of the sawing machine spindle motor at various times; obtain the inherent response delay of the sawing machine through system calibration; obtain the reference current of the spindle motor, the reference speed of the sawing machine, the sampling time interval of the CNC system, the overhang length of the saw band, the material thickness of the current workpiece being cut, and the maximum safe skew displacement threshold of the sawing machine to construct the basic data source.

[0025] It should be noted that when a sawing machine is machining complex metal parts at high speed, the load fluctuations of the spindle motor and the mechanical deformation of the saw blade directly affect the final cutting accuracy. To comprehensively understand the real-time cutting status of the sawing machine, this invention uses multi-modal sensors to jointly collect the real-time operating parameters of the sawing machine. Furthermore, to subsequently convert the transient mechanical fluctuations of the sawing machine into accurate compensation commands, it is also necessary to extract benchmark thresholds related to the physical structure of the sawing machine and the machining target in advance, thereby providing complete underlying data support for high-precision intelligent control.

[0026] Specifically, the current, feed speed, feed acceleration, skew displacement, and torsional deformation angle of the saw spindle motor at various times are obtained, including: It should be noted that, through current and displacement sensors, minute dynamic changes in the transmission chain and actuator of the saw can be captured in real time. Since the data acquisition of the CNC system has discrete characteristics, derivative calculations are performed using the difference between the saw's time points to extract the physical rate of change; the sampling frequencies of all sensors are strictly synchronized with the sampling frequency of the CNC system to ensure consistent data timing.

[0027] The current of the spindle motor is collected in real time by a current sensor installed on the spindle motor of the saw; the feed speed of the saw is collected in real time by a photoelectric encoder installed on the feed drive shaft of the saw; and the feed acceleration of the saw at each moment is obtained based on the ratio of the difference in feed speed between adjacent moments of the saw to the sampling time interval of the CNC system.

[0028] The skew displacement of the saw blade at various times is obtained by a displacement sensor installed on the guide arm of the sawing machine, and the torsional deformation angle of the saw blade at various times is obtained by differential calculation using data from dual displacement sensors.

[0029] Preferably, the inherent system response delay of the saw is obtained through system calibration, including: It should be noted that the inherent response delay of the system refers to the fixed physical delay from the issuance of control commands by the CNC system to the effective action of the feed system actuator. It serves as the benchmark for subsequent spatiotemporal misalignment analysis of the CNC system and includes data transmission time, algorithm calculation time, and hydraulic hardware execution time.

[0030] The inherent system response latency of the saw, including data transmission time and average algorithm calculation time, is tested and obtained through the saw's system calibration module.

[0031] Preferably, the basic data source is constructed by acquiring the reference current of the spindle motor, the machine tool reference speed of the saw, the sampling time interval of the CNC system, the overhang length of the saw blade, the material thickness of the current workpiece being cut, and the maximum safe skew displacement threshold of the saw, including: It should be noted that by introducing the fixed parameters of the sawing machine and the current size of the workpiece being cut, the dimensionless sensing characteristics can be mapped to the real physical processing space.

[0032] The reference current of the spindle motor is obtained through a no-load operation test of the saw; the machine tool reference speed calibrated at the factory is also obtained. For example, 50% of the rated maximum feed speed of the saw is used as the machine tool reference speed.

[0033] Retrieve the sampling time interval set by the CNC system. For example, the sampling time interval of the CNC system is set to 0.01s. If the sampling time interval of the CNC system is too large, such as 0.5s, it will cause the CNC system to miss the high-frequency transient change characteristics of the spindle motor; if the sampling time interval of the CNC system is too small, such as 0.001s, it will cause the CNC system to be filled with high-frequency sensor noise, increasing the hardware computing power burden of the CNC system.

[0034] The length of the free section between the two guide arms of the saw blade is obtained through the machine tool parameter reading module and recorded as the saw blade suspension length; the material thickness of the workpiece being cut is obtained through the machine tool parameter setting interface.

[0035] Obtain the maximum safe deviation displacement threshold set by the CNC system for the saw. For example, the maximum safe deviation displacement threshold is set to 2mm. If the maximum safe deviation displacement threshold is too large, such as 5mm, the saw will not be able to trigger strong control before cutting out severely out-of-tolerance scrap; if the maximum safe deviation displacement threshold is too small, such as 0.1mm, the CNC system will be overly sensitive to normal slight process vibrations, causing frequent vibrations in the saw's actuator.

[0036] The spindle motor current, the saw feed speed, the current feed acceleration of the saw, the current skew displacement of the saw band, the current torsional deformation angle of the saw band, the inherent system response delay of the saw, the reference current of the spindle motor, the machine tool reference speed of the saw, the sampling time interval of the CNC system, the saw band overhang length, the material thickness of the current workpiece being cut, and the maximum safe skew displacement threshold of the saw together constitute the basic data source for controlling cutting accuracy.

[0037] At this point, the basic data source for adjusting cutting precision has been obtained.

[0038] S2: Obtain the dynamic hysteresis time of the CNC sawing machine. The specific steps are as follows: Based on the current difference between adjacent moments of the spindle motor, obtain the resistance change characteristic of the sawing process; based on the relative change characteristics of the feed acceleration of the sawing machine, obtain the inertial response characteristic of the feed system; based on the saw band suspension length and the feed speed of the sawing machine, combined with the relative hysteresis effect of the resistance change characteristic on the inertial response characteristic, obtain the dynamic hysteresis time of the CNC sawing machine.

[0039] It should be noted that when a saw is cutting large metal parts, the real-time changes in the spindle motor current at various moments directly reflect the abrupt changes in cutting resistance during the sawing process, while the current feed acceleration of the saw's feed system reflects the mechanical response capability of the saw's underlying components. Since the transmission of abrupt changes in cutting resistance to the feed system requires overcoming the mass inertia of the transmission chain, the asymmetry between the abrupt changes in cutting resistance and the mechanical response of the feed system on the time axis inevitably leads to a dynamic hysteresis effect within the saw blade currently cutting the workpiece. Therefore, this invention quantifies the degree of mechanical response hysteresis caused by cutting resistance by extracting and comparing the characteristics of the abrupt changes in spindle motor current and the response of the motion system.

[0040] Specifically, based on the current difference between adjacent moments of the spindle motor, the resistance change characteristic of the sawing process is obtained; based on the relative change characteristics of the saw's feed acceleration, the inertial response characteristic of the feed system is obtained, including: It should be noted that the increase in cutting resistance is reflected in the rise of the spindle current of the spindle motor. When constructing the resistance mutation characterization term in the sawing process, the theoretical basis of this invention comes from the well-known linear mapping principle of motor torque and current intensity in the field of electromechanical control. This invention uses the algebraic division ratio of the absolute difference of current at adjacent moments and the reference current, combined with the sensitivity constant, to directly generate a purely linear dimensionless amplification term, which maps the sudden oscillation level of cutting resistance.

[0041] The preset resistance change sensitivity constant is denoted as: This is used to adjust the system's response gain to sudden current changes. For example, Set to 2. If If the value is too large, such as 10, it will cause the resistance mutation characteristic term to be artificially high, and the CNC system will over-stress the normal material texture changes; if If the value is too small, such as 0.1, it will cause the system to become passive, ignoring the severe resistance spikes caused by inclusion hard points.

[0042] The resistance mutation characterization term in the sawing process satisfies the expression: ; In the formula, The term representing the sudden change in resistance at time t during the sawing process is dimensionless. , This represents the current of the spindle motor at time t and time t-1; This indicates the reference current of the spindle motor; Represents the absolute value function; This represents the resistance change sensitivity constant.

[0043] In the formula, This represents the relative volatility of the spindle motor, used to express the degree of relative change in the current of the spindle motor between adjacent sampling times; It represents the net change in resistance after being amplified by the sensitivity constant; the whole is superimposed on the basic constant 1 to form the resistance change characterization term of the pure algebraic sawing process, which intuitively reflects the relative multiple of the sharp increase in cutting resistance.

[0044] It should be noted that the inertial resistance and physical response of the feed system depend not only on the current feed acceleration of the feed system and the sampling time interval of the CNC system, but also on the normalization process using the machine tool reference speed of the CNC sawing machine. The basis for constructing the inertial response characterization term of the feed system in this invention is derived from the well-known kinematic principles of Newton's second law and the momentum theorem in physical kinematics, used for approximate evaluation. This invention divides the velocity increment obtained by integrating the acceleration with the reference speed to restore the impact resistance and physical damping capability of the mechanical chassis.

[0045] The preset inertial response sensitivity constant is denoted as... For example, Set to 1.

[0046] The inertial response characterization term of the feed system satisfies the following expression: ; In the formula, The dimensionless term represents the inertial response of the feed system at time t. This represents the feed acceleration of the saw at time t. Indicates the sampling time interval of the CNC system; Indicates the machine tool reference speed of the saw. This represents the preset inertial response sensitivity constant.

[0047] In the formula, This represents the acceleration response characteristic value of the feed system after dimensionless processing; This represents the underlying physical response characteristics constructed based on the natural exponential function; This represents the inertial response characterization term of the feed system constructed by combining the fundamental constant and the sensitivity constant. The larger this value is, the stronger the resistance of the feed system to drag disturbances and the weaker the hysteresis effect.

[0048] Preferably, the sawing dynamic hysteresis time of the CNC sawing machine is obtained based on the saw band overhang length and the feed speed of the sawing machine, combined with the relative hysteresis effect of the resistance change characterization term on the inertial response characterization term, including: It should be noted that the longer the saw band is suspended and the slower the real-time feed speed of the sawing machine, the greater the response hysteresis time span of the sawing action under forced vibration and resistance disturbance during sawing, and the more significant the sawing hysteresis effect. Based on this fundamental law of mechanical dynamics, this invention models the dynamic hysteresis time of CNC sawing. The basic force time span of the saw band, composed of the suspended length of the saw band and the real-time feed speed, is used as the hysteresis calculation benchmark. The hysteresis effect is positively corrected by the resistance mutation characterization term of the sawing process, and the hysteresis effect is negatively constrained by the inertial response characterization term of the feed system. The three are combined to construct the dynamic hysteresis time of CNC sawing under all working conditions.

[0049] The dynamic hysteresis time of a CNC sawing machine satisfies the expression: ; In the formula, The dynamic lag time of the CNC saw at time t is expressed in time and the unit is seconds. The term representing the sudden change in resistance at time t during the sawing process; This represents the inertial response characteristic of the feed system at time t. Indicates the length of the saw blade suspended in the air; This represents the feed rate of the saw at time t. This indicates a preset small positive value to avoid a denominator of 0. For example... The unit is consistent with the feed rate unit.

[0050] In the formula, This indicates the relative hysteresis effect after the sudden change in resistance is weakened by the inertial damping of the feed system, reflecting the proportional relationship of the actual conversion of cutting resistance into mechanical response hysteresis. This represents the basic stress time span term of the saw blade, which is composed of the saw blade overhang length and the safe feed speed of the sawing machine. It represents a composite calculation term that combines the sudden change in combined resistance, system inertia, and the time of force application, and evaluates the dynamic hysteresis level during the sawing process through product mapping.

[0051] It should be noted that, as Figure 2 The graph shows the dynamic hysteresis time of sawing over time. The curve in the graph represents the dynamic change trend of the dynamic hysteresis time of sawing during the cutting process of the CNC sawing machine. It shows the ratio of the resistance mutation characteristic to the inertial response characteristic of the feed system during sawing. The dynamic hysteresis level calculated by combining the saw blade suspension length and feed speed provides a basis for the subsequent calculation of the feedforward correction gain coefficient.

[0052] Thus, the dynamic hysteresis time of the CNC sawing machine was obtained.

[0053] S3: Obtain the feedforward correction gain coefficient of the CNC system. The specific steps are as follows: amplify the inherent response delay of the system by combining the sawing dynamic hysteresis time, and project the sliding distance by the feed speed and torsional deformation angle of the saw to obtain the spatiotemporal misalignment compensation span of the saw; according to the offset ratio of the spatiotemporal misalignment compensation span relative to the material thickness of the current workpiece, and combined with the relative skew displacement of the saw band, obtain the feedforward correction gain coefficient of the CNC system.

[0054] It should be noted that due to the inherent physical time delay in the calculation of the control algorithm and the execution of the hydraulic hardware, the saw band continues to cut within the metal at the current feed speed of the sawing machine during the inherent system response delay. Considering the dynamic hysteresis time of the CNC sawing machine, the actual length of the erroneous trajectory traversed by the saw band during this inherent system response delay is not the product of the current feed speed of the sawing machine and the inherent system response delay, but rather is subject to a strong nonlinear stretching due to the metal resistance effect. Furthermore, the physical characteristics of the saw band's tension recovery under different lateral skew displacements conform to a nonlinear extension law. Therefore, this invention first evaluates the absolute amount of the saw band's spatiotemporal offset, and then, combined with the saw band's skew characteristics, converts the absolute amount into a feedforward correction gain coefficient for the CNC system, thereby performing spatial and temporal alignment corrections on the control commands.

[0055] Specifically, the inherent response delay of the system is amplified by combining the dynamic hysteresis time of the sawing, and the spatiotemporal misalignment compensation span of the saw is obtained by projecting the sliding distance using the feed speed and torsional deformation angle of the saw, including: It should be noted that the inherent response delay of the system represents the control gap period that the control commands cannot reach. To accurately calculate the physical offset of the saw band during this control gap period, this invention superimposes the dynamic hysteresis time of the sawing process onto the inherent response delay of the system to form the total hysteresis time. Then, a relative misalignment amplification multiplier is constructed using the ratio of this total hysteresis time to the inherent response delay. The basic linear sliding distance within the inherent response delay is calculated using the real-time feed rate. Based on the Taylor series first-order expansion approximation theorem and the geometrical linear attenuation projection model, this invention directly uses the linear subtraction of the absolute value of the angle and the attenuation constant under conditions of small torsional deformation to achieve algebraic projection reduction.

[0056] The preset torsional attenuation projection constant is denoted as... Exemplary Set to 0.5 / rad. If the value is too large, the projection will be excessively reduced; if the value is too small, it will not reflect the decrease in axial cutting capability caused by torsion.

[0057] The spatiotemporal misalignment compensation span of the saw satisfies the expression: ; In the formula, The span of the spatiotemporal misalignment compensation of the saw at time t is expressed in length and the unit is millimeters. This represents the feed rate of the saw at time t. This represents the inherent system response delay of the saw; This represents the dynamic hysteresis time of the CNC saw at time t. The angle of torsional deformation of the saw blade at time t is expressed in radians. This represents the torsional attenuation projection constant.

[0058] In the formula, This represents the basic linear sliding distance of the saw at time t within the inherent response time delay of the system; This represents the relative misalignment amplification multiplier caused by the degree of mechanical response lag; This represents the pure algebraic effective physical projection ratio of the saw band's torsional posture in the cutting feed direction; This reflects the actual physical offset distance of the saw band during the inherent system response delay of the CNC sawing machine, caused by the combination of dynamic hysteresis and torsional posture.

[0059] Preferably, the feedforward correction gain coefficient of the CNC system is obtained based on the offset ratio of the spatiotemporal misalignment compensation span relative to the material thickness of the currently cut workpiece, combined with the relative skew displacement of the saw blade, including: It should be noted that the spatiotemporal misalignment compensation span is a physical distance scale. To convert it into a dimensionless multiplicative adjustment factor for the CNC system's drive signal, it is necessary to calculate the relative proportion between the spatiotemporal misalignment compensation span and the current cutting workpiece material thickness. Simultaneously, to avoid excessive feedforward compensation strength causing saw blade sway to exceed the safety threshold and leading to equipment damage, nonlinear constraints must be applied to the adjustment factor. The theoretical basis for the formula in this invention comes from the well-known inverse proportional saturation nonlinearity suppression and linear normalization penalty laws in the field of automatic control.

[0060] The feedforward correction gain coefficient of the CNC system satisfies the following expression: ; In the formula, The feedforward correction gain coefficient of the CNC system at time t is dimensionless. This represents the span of the spatiotemporal misalignment compensation for the saw at time t. Indicates the material thickness of the workpiece currently being cut; The absolute value of the skew displacement of the saw blade at time t is represented by a scalar. This indicates the maximum safe skew displacement threshold for the band saw. It should be noted that... , , , All of these have the dimension of length, and the unit is millimeter.

[0061] In the formula, The term represents the inverse proportional offset saturation term of the spatiotemporal misalignment compensation span of the sawing machine relative to the thickness of the material of the workpiece being cut. Since the denominator includes the numerator, its upper limit is naturally and strictly locked to a limit domain below 1, avoiding instability of the compensation strength under extreme working conditions, and the dimensions are completely canceled. This represents a purely linear tension constraint penalty term consisting of the absolute value scalar of the saw blade's skew displacement at time t and the ratio of the maximum safe skew displacement threshold of the CNC saw. This means that the compensation multiplier is superimposed on the basic constant 1 after the two dimensionless proportional terms are merged, so as to ensure that the feedforward correction gain coefficient of the obtained CNC system at time t is always greater than or equal to 1, and only the lower-level drive signal is positively and safely amplified and compensated.

[0062] It should be noted that, as Figure 3 The graph shows the change of the feedforward correction gain coefficient of the CNC system over time. The curve in the graph represents the dynamic change trend of the feedforward correction gain coefficient over time during the cutting process of the CNC sawing machine. It shows the distribution characteristics of the feedforward correction gain coefficient and reflects the intensity of spatial and temporal alignment correction of the basic compensation amount at different times.

[0063] Thus, the feedforward correction gain coefficient of the CNC system at any time has been obtained.

[0064] S4: To achieve intelligent control of the sawing machine's cutting accuracy, the specific steps are as follows: Input the current, feed speed, feed acceleration, skew displacement, and torsional deformation angle of the spindle motor into the deep learning model, output the basic feed speed compensation amount and basic tension compensation amount of the sawing machine, and adjust them using the feedforward correction gain coefficient to obtain the final feed speed compensation amount and final tension compensation amount of the sawing machine, thereby achieving intelligent control of the sawing machine's cutting accuracy.

[0065] It should be noted that during heavy-duty cutting, the multimodal sensor signals of a sawing machine contain complex nonlinear time-dependent characteristics, making it difficult for traditional linear control algorithms to effectively decouple and predict multiple physical parameters. Therefore, this invention utilizes the powerful time-series analysis capabilities of deep learning models to extract the basic feed rate compensation and basic tension compensation of the saw blade for the current cutting conditions. Subsequently, the feedforward correction gain coefficients of the CNC system obtained from the aforementioned calculations are used for damping scaling and amplification enhancement, respectively, and finally sent to the underlying actuator of the sawing machine to complete the intelligent control of the cutting.

[0066] Specifically, the current, feed speed, feed acceleration, skew displacement, and torsional deformation angle of the spindle motor are input into the deep learning model, which outputs the basic feed speed compensation and basic tension compensation of the saw. These are then adjusted using the feedforward correction gain coefficient to obtain the final feed speed compensation and final tension compensation of the saw, thus achieving intelligent control of the saw's cutting accuracy. This includes: It should be noted that deep learning algorithms possess excellent capabilities for extracting historical sequence features. To adapt to dynamic and continuous metal cutting scenarios, a reasonable time step window must be set to extract continuous sensing features for the entire working cycle.

[0067] Set the historical time window length for the deep learning model. For example, set the historical time window length for the pre-trained deep learning model to 30 time steps. If the historical time window length of the pre-trained deep learning model is too long, such as 100 time steps, it will introduce too many outdated processing states, causing the pre-trained deep learning model to be sluggish. If the historical time window length of the pre-trained deep learning model is too short, such as 5 time steps, it will not be able to fully cover a complete mechanical vibration cycle of the saw, causing the basic feed rate compensation output by the pre-trained deep learning model to fluctuate frequently and drastically.

[0068] A continuous multi-source sensing data sequence, including the current of the spindle motor at various times, the feed rate, the feed acceleration, the skew displacement, and the torsional deformation angle, is input into the deep learning model. For example, the deep learning model employs a Long Short-Term Memory (LSTM) network. Its internal forget gate and input gate control mechanism effectively filters out redundant process noise and retains long-term machine tool deformation trend characteristics. The model network structure includes three LSTM hidden layers, with 128 nodes per hidden layer. The input layer dimension is 30×5, with 30 time steps, each time step containing 5-dimensional features. The output layer dimension is 2, corresponding to the basic feed rate compensation and the basic tension compensation, respectively.

[0069] The basic feed rate compensation and basic tension compensation of the saw blade are calculated by forward inference through a deep learning model, and the output is based on the current cutting conditions.

[0070] It should be noted that, in order to achieve rapid alignment between physical space and control commands, the basic feed speed compensation of the saw output by the deep learning model needs to be quickly reduced when the misalignment is severe, so as to allow time for the saw band to recover from deformation. Therefore, the feed speed compensation adopts division damping adjustment logic. On the other hand, the basic tension compensation of the saw band output by the deep learning model needs to output a larger compensation force when the misalignment span increases. Therefore, the tension compensation adopts multiplication amplification enhancement logic to completely eliminate hysteresis and forcibly straighten the saw band.

[0071] The final feed rate compensation of the saw satisfies the expression: ; In the formula, This represents the final feed rate compensation amount of the saw at time t. This indicates the basic feed rate compensation amount of the banding machine; This represents the feedforward correction gain coefficient of the CNC system at time t.

[0072] In the formula, This represents the terminal correction calculation process that uses the feedforward correction gain coefficient of the CNC system at time t to apply speed reduction damping adjustment to the feed speed command. This allows the servo drive of the saw to quickly reduce the feed speed of the saw when the saw band is severely skewed, leaving sufficient time margin for the elastic reset of the saw band.

[0073] The final tension compensation of the saw belt satisfies the following expression: ; In the formula, This represents the final tension compensation amount of the saw belt at time t; This represents the compensation amount of the basic tension force of the saw belt at time t; This represents the feedforward correction gain coefficient of the CNC system at time t.

[0074] In the formula, This refers to the terminal correction calculation process that uses the feedforward correction gain coefficient of the CNC system at time t to amplify and enhance the tension command, so that the hydraulic control valve of the sawing machine outputs an additional tensile force multiple times when the spatiotemporal misalignment compensation span of the sawing machine increases, thereby rigidly suppressing the deformation and deviation of the saw band.

[0075] The final feed rate compensation of the saw is sent to the servo drive of the saw, and the calculated final tension compensation of the saw band is sent to the hydraulic control valve of the saw, so as to calibrate the servo drive and the hydraulic control valve of the saw.

[0076] This completes the intelligent control of the saw's cutting precision.

[0077] This invention also discloses an intelligent control system for the cutting precision of a large CNC cutting saw, comprising a processor and a memory. The memory stores computer program instructions, and when the computer program instructions are executed by the processor, the intelligent control method for the cutting precision of a large CNC cutting saw according to this invention is realized.

[0078] The system also includes other components well known to those skilled in the art, such as communication buses and communication interfaces, the settings and functions of which are known in the art and will not be described in detail here.

[0079] While this specification has shown and described numerous embodiments of the invention, it will be apparent to those skilled in the art that such embodiments are provided by way of example only. Many modifications, alterations, and alternatives will occur to those skilled in the art without departing from the spirit and essence of the invention. It should be understood that various alternatives to the embodiments of the invention described herein may be employed in the practice of this invention.

Claims

1. A method for intelligent control of cutting accuracy on a large CNC cutting saw, characterized in that, include: The system acquires the current and reference current of the spindle motor of the sawing machine at various times, the feed speed of the sawing machine, the feed acceleration of the sawing machine, the skew displacement and torsional deformation angle of the saw blade; it also acquires the inherent response delay of the sawing machine through system calibration, the suspended length of the saw blade, the material thickness of the workpiece being cut, the maximum safe skew displacement threshold of the sawing machine, and the sampling time interval of the CNC system and the machine tool reference speed of the sawing machine. Based on the current difference between adjacent moments of the spindle motor and the feed acceleration of the saw, combined with the saw band suspension length and the feed speed of the saw, the sawing dynamic hysteresis time of the CNC saw is obtained. The inherent response delay of the system is amplified by combining the dynamic hysteresis time of the sawing process, and the spatiotemporal misalignment compensation span of the saw is obtained by projecting it onto the feed speed and torsional deformation angle of the saw. Based on the offset ratio of the spatiotemporal misalignment compensation span relative to the material thickness of the currently cut workpiece, and combined with the skew displacement of the saw band, the feedforward correction gain coefficient of the CNC system is obtained. The current of the spindle motor, the feed speed of the saw, the feed acceleration of the saw, the skew displacement of the saw band, and the torsional deformation angle of the saw band are input into the deep learning model, and the basic feed speed compensation amount and basic tension compensation amount of the saw are output. The feedforward correction gain coefficient is used for adjustment to obtain the final feed speed compensation amount and final tension compensation amount of the saw.

2. The intelligent control method for cutting accuracy of a large CNC cutting saw as described in claim 1, characterized in that, The process of obtaining the dynamic hysteresis time of the CNC saw includes: Obtain the resistance mutation characteristic of the sawing process; obtain the inertial response characteristic of the feed system; multiply the ratio of the resistance mutation characteristic at any time to the inertial response characteristic at that time by the product of the suspended length of the saw band and the feed speed of the sawing machine at that time to obtain the sawing dynamic hysteresis time of the CNC sawing machine at that time.

3. The intelligent control method for cutting accuracy of a large CNC cutting saw as described in claim 1, characterized in that, The spatiotemporal misalignment compensation span satisfies the expression: ； In the formula, This represents the span of the spatiotemporal misalignment compensation for the saw at time t. This represents the feed rate of the saw at time t. This represents the inherent system response delay of the saw; This represents the dynamic hysteresis time of the CNC saw at time t. This represents the torsional deformation angle of the saw blade at time t. This represents the torsional attenuation projection constant.

4. The intelligent control method for cutting accuracy of a large CNC cutting saw as described in claim 2, characterized in that, The resistance mutation characterization term satisfies the expression: ； In the formula, The term representing the sudden change in resistance at time t during the sawing process; , This represents the current of the spindle motor at time t and time t-1; This indicates the reference current of the spindle motor; Represents the absolute value function; This represents the resistance change sensitivity constant.

5. The intelligent control method for cutting accuracy of a large CNC cutting saw as described in claim 2, characterized in that, The inertial response characterization term of the feed system satisfies the expression: ； In the formula, This represents the inertial response characteristic of the feed system at time t. This represents the feed acceleration of the saw at time t. Indicates the sampling time interval of the CNC system; This indicates the machine tool reference speed of the saw; This represents the preset inertial response sensitivity constant.

6. The intelligent control method for cutting accuracy of a large CNC cutting saw as described in claim 1, characterized in that, The feedforward correction gain coefficient of the CNC system satisfies the following expression: ； In the formula, This represents the feedforward correction gain coefficient of the CNC system at time t; This represents the span of the spatiotemporal misalignment compensation for the saw at time t. Indicates the material thickness of the workpiece currently being cut; The absolute value of the skew displacement of the saw blade at time t is represented by a scalar. This indicates the maximum safe skew displacement threshold for the saw.

7. The intelligent control method for cutting accuracy of a large CNC cutting saw as described in claim 1, characterized in that, The adjustment using the feedforward correction gain coefficient includes: The ratio of the basic feed rate compensation of the saw to the feedforward correction gain coefficient is used as the final feed rate compensation of the saw. The product of the basic tension compensation amount and the feedforward correction gain coefficient is used as the final tension compensation amount of the saw belt.

8. The intelligent control method for cutting accuracy of a large CNC cutting saw as described in claim 1, characterized in that, The acquisition of the current and reference current of the saw spindle motor at various times, the feed speed of the saw, the feed acceleration of the saw, the skew displacement of the saw band, and the torsional deformation angle include: The current of the main spindle motor is collected by a current sensor installed on the main spindle motor of the saw; the reference current of the main spindle motor is obtained by the no-load operation test of the saw; the feed speed of the saw is collected by an optical encoder installed on the feed drive shaft of the saw; the feed acceleration of the saw at each moment is obtained based on the ratio of the difference in feed speed at adjacent moments to the sampling time interval; the skew displacement of the saw band at each moment is obtained by a displacement sensor installed on the guide arm of the saw, and the torsional deformation angle of the saw band at each moment is obtained by differential calculation of dual-channel sensor data.

9. The intelligent control method for cutting accuracy of a large CNC cutting saw according to claim 1, characterized in that, The acquisition of the inherent response delay of the sawing system, the acquisition of the saw blade overhang length, the material thickness of the currently cut workpiece, the maximum safe skew displacement threshold of the sawing, and the sampling time interval of the CNC system and the machine tool reference speed of the sawing include: The inherent response delay of the system is obtained through the system calibration module of the saw; the overhead length of the saw band, the sampling time interval of the CNC system, and the machine tool reference speed of the saw are obtained through the machine tool parameter reading module; the material thickness of the workpiece being cut is obtained through the machine tool parameter setting interface; and the maximum safe deviation displacement threshold of the saw set by the CNC system is obtained.

10. A large-scale CNC cutting sawing machine cutting precision intelligent control system, characterized in that, include: A processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, a method for intelligent control of cutting accuracy of a large CNC cutting saw as described in any one of claims 1-9 is provided.

Images