Anti-jamming chain type automatic feeding of nail industrial control system and method

Abstract

The application discloses a kind of anti-stuck chain belt type automatic feeding industrial control system and method, belong to machine tool fieldbus control system technical field, comprising: in operation high-frequency perturbation excitation signal is applied, and multi-dimensional state signals such as motor current, rotating speed and vibration acceleration are collected;Dynamic mechanical impedance spectrum is identified based on electromechanical coupling model, and impedance eigenvector representing microscopic stick-slip effect is extracted;According to the vector, the resistance evolution trajectory is predicted and the risk level of sticking is determined;When there is risk, impedance parameters are used as hard constraint conditions, and a nonlinear speed-torque composite control trajectory is generated based on multi-objective optimization;Finally, the adaptive anti-sticking action is executed by the driving system and the model is iteratively corrected.The application breaks through the hysteresis of traditional threshold monitoring, accurately captures the micro-sticking precursor, realizes active defense, and effectively guarantees the production rhythm.

Description

Technical Field

[0001] This invention relates to the field of machine tool fieldbus control system technology, and more particularly to an anti-jamming chain-type automatic nail feeding industrial control system and method. Background Technology

[0002] Chain-driven automatic fastener feeding systems are widely used in high-end equipment manufacturing fields such as aerospace, automotive manufacturing, and precision electronics assembly. They are primarily used to continuously and stably transport various fasteners to assembly stations. This system typically consists of a drive motor, a chain conveyor mechanism, guide rails, and a control unit. The intermittent or continuous movement of the chain drives the fasteners along a predetermined trajectory. In automated production lines, the stability of the fastener feeding system directly determines the overall production cycle time and assembly quality. With the development of intelligent manufacturing technology, modern fastener feeding equipment is generally equipped with basic electrical control modules, enabling functions such as motor start / stop, speed adjustment, and simple fault alarms. Under standard operating conditions, such systems can meet basic material feeding requirements. Their working principle mainly relies on the coordination of the mechanical structure and preset electrical logic, maintaining operation by detecting the chain position or motor load status through sensors. However, in actual high-cycle, high-precision production scenarios, the fastener feeding process faces complex changes in tribological properties, fluctuations in material dimensional tolerances, and nonlinear interference caused by mechanical wear.

[0003] Existing chain-type automatic nail feeding control systems with anti-jamming functions typically employ detection mechanisms based on single or a few physical quantity thresholds to handle anomalies. These systems often monitor the current amplitude of the drive motor or the chain speed in real time. When the current exceeds a preset upper limit or the speed falls below a set lower limit, a jamming or pre-jamming phenomenon is identified. Once the threshold is triggered, the control system executes a preset linear deceleration, short-term stop, or fixed-frequency reciprocating jitter strategy to attempt to eliminate the resistance anomaly. Some advanced solutions introduce vibration sensors, performing simple weighted fusion of vibration amplitude and current signals to improve detection sensitivity. However, these existing technologies are essentially still rule-based open-loop or simple closed-loop feedback control, lacking in-depth analysis of the system's internal electromechanical coupling characteristics. Their drawback lies in their inability to distinguish between false anomalies caused by voltage fluctuations and instantaneous load impacts and true jamming precursors caused by microscopic stick-slip effects, resulting in a high false alarm rate. Frequent unnecessary decelerations severely reduce production cycle time, and they often fail when facing complex and variable jamming conditions. Summary of the Invention

[0004] This invention overcomes the shortcomings of the prior art and provides an anti-jamming chain-type automatic nail feeding industrial control system and method.

[0005] To achieve the above objectives, the technical solution adopted by the present invention is: an anti-jamming chain-type automatic nail feeding industrial control system, comprising: Chain-driven fastener feeding mechanism for carrying and feeding fasteners; A servo drive unit, connected to the chain-type nail feeding mechanical actuator, is used to provide driving force and has the function of switching between speed control mode and torque control mode; The multi-source sensing unit includes sensors for acquiring current and speed signals of the drive motor and sensors for acquiring vibration acceleration signals of the chain drive mechanism. The programmable logic controller is communicatively connected to the servo drive unit and the multi-source sensing unit, and has a built-in perturbation excitation signal generator, a multi-source signal synchronous acquisition module and a frequency division domain decoupling drive module. It is used to generate high-frequency perturbation excitation signals, synchronously acquire multi-dimensional state signals, and decompose the composite control trajectory and distribute it to the speed loop and torque loop of the servo drive unit. An industrial control computer is connected to the programmable logic controller and has built-in dynamic impedance spectrum identification module, impedance feature vector extraction module, resistance evolution trajectory prediction module, jamming risk level determination module, nonlinear speed-torque composite trajectory optimization module, and adaptive closed-loop correction module. The industrial control computer is configured to: identify the dynamic mechanical impedance spectrum and extract the impedance feature vector based on the multidimensional state signal using an electromechanical coupling mathematical model; predict the resistance evolution trajectory and determine the jamming risk level based on the impedance feature vector; when a jamming risk is determined, generate a nonlinear speed-torque composite control trajectory based on a multi-objective optimization function and combined with the real-time identified equivalent stiffness coefficient and equivalent damping coefficient; and iteratively correct the electromechanical coupling mathematical model based on the actual system response feedback.

[0006] In a preferred embodiment of the present invention, the perturbation excitation signal generator in the programmable logic controller is configured to superimpose an additional electrical instruction with limited amplitude and controlled frequency on the basic operating instruction of the drive motor. The additional electrical instruction is one of a multi-frequency simple harmonic wave superposition sequence, a sinusoidal sweep signal, or a pseudo-random binary sequence, and its frequency band covers a wide range from low-frequency structural modes to high-frequency contact resonance peaks. The multi-source signal synchronous acquisition module is configured to use the encoder zero pulse signal of the servo driver as the main clock reference, and to use the hardware triggering unit to perform nanosecond-level synchronous acquisition of current signal, rotation speed signal and vibration acceleration signal, and to attach a high-precision timestamp to each data point to form a multi-dimensional state signal vector.

[0007] In a preferred embodiment of the present invention, the nonlinear speed-torque composite trajectory optimization module is configured to adopt a nonlinear model predictive control framework, construct a state-space model including the motor dynamics equation, the transmission chain elastic deformation equation and the friction evolution equation as a prediction model, embed the real-time identified equivalent stiffness coefficient and equivalent damping coefficient as hard constraints into the state-space model, and generate a nonlinear speed-torque composite control trajectory including dynamically adjusted oscillation frequency, amplitude and phase sequence by solving a multi-objective optimization function under the premise of satisfying the motor's maximum current, maximum speed and mechanical strength constraints. The frequency-domain decoupling drive module is configured to decompose the nonlinear speed-torque composite control trajectory into a low-frequency fundamental component and a high-frequency micro-vibration component using a filter bank with an adjustable cutoff frequency. The low-frequency fundamental component is mapped to the speed control loop of the servo drive unit, and the high-frequency micro-vibration component is mapped to the torque control loop of the servo drive unit.

[0008] This invention provides an industrial control method for preventing jamming of a chain-driven automatic nail feeding system, comprising the following steps: S1. During the operation of the chain-type automatic nail feeding system, a high-frequency micro-perturbation excitation signal is applied, and multi-dimensional status signals under the system operation status are collected, including the real-time current and speed of the drive motor, and the vibration acceleration of the chain-driven mechanism. S2. Based on the electromechanical coupling mathematical model, the multidimensional state signal is analyzed to identify the dynamic mechanical impedance spectrum of the contact interface between the chain and the nail, and the impedance feature vector characterizing the evolution trend of the micro-stick-slip effect is extracted from it. S3. Predict the resistance evolution trajectory of the nail supply system based on the impedance characteristic vector, and then determine the current jamming risk level; S4. When it is determined that there is a risk of jamming, a nonlinear velocity-torque composite control trajectory adapted to the current physical characteristics of the system is generated based on a multi-objective optimization function. S5. Based on the nonlinear speed-torque composite control trajectory drive system, perform adaptive anti-jamming actions, and iteratively correct the electromechanical coupling mathematical model according to the actual system response.

[0009] In a preferred embodiment of the present invention, the process of applying a high-frequency perturbation excitation signal in step S1 includes: Additional electrical commands are superimposed on the basic operating commands of the drive motor. The amplitude of the additional electrical commands is limited to a threshold that does not interfere with the normal nail feeding cycle and conveying accuracy, and its frequency band covers a wide range from low-frequency structural modes to high-frequency contact resonance peaks. While applying the high-frequency perturbation excitation signal, the real-time current and speed of the drive motor and the vibration acceleration of the chain conveyor mechanism are synchronously acquired in time to form a multi-dimensional state signal vector.

[0010] In a preferred embodiment of the present invention, the process of extracting the impedance feature vector in step S2 includes: The stiffness coefficient and damping coefficient can be identified in real time. The stiffness coefficient and damping coefficient within the preset time window are transformed in the frequency domain to construct a dynamic mechanical impedance spectrum; The resonant frequency offset, damping loss factor peak value, and spectral energy distribution entropy are selected from the dynamic mechanical impedance spectrum to form the impedance characteristic vector.

[0011] In a preferred embodiment of the present invention, step S3, predicting the resistance evolution trajectory and determining the risk level, includes: The impedance characteristic vector sequence within the historical period is mapped to a high-dimensional phase space to construct the mapping relationship between impedance parameters and future resistance values; A continuous curve of resistance changing with time over a future period can be obtained by numerical integration. The morphological characteristics of the continuous curve and the probability distribution of the predicted resistance peak exceeding the system driving threshold are calculated, and a risk index function is constructed in combination. The risk index function is quantized and mapped to multiple preset risk level ranges.

[0012] In a preferred embodiment of the present invention, generating the nonlinear velocity-torque composite control trajectory in step S4 includes: The equivalent stiffness coefficient and equivalent damping coefficient identified in real time are read as internal time-varying parameters of the control model, and the optimized boundary conditions are determined in combination with the determined risk level. A state-space model containing the motor dynamics equations, the transmission chain elastic deformation equations, and the friction evolution equations is constructed as a prediction model. Using a model predictive control framework, a set of speed and torque control sequences that minimize the multi-objective optimization function is searched within a future finite time domain. The multi-objective optimization function includes four sub-objectives: minimizing the probability of drag overshoot, minimizing the speed tracking error, minimizing the jerk, and minimizing the control energy consumption. The weight coefficients of each sub-objective are adjusted in real time according to the risk level.

[0013] In a preferred embodiment of the present invention, the adaptive de-jamming action performed in step S5 specifically includes: A frequency-domain decoupling strategy is adopted to decompose the nonlinear velocity-torque composite control trajectory into a low-frequency fundamental component and a high-frequency micro-vibration component; The low-frequency fundamental component is sent into the motor's speed control loop as a main speed reference command. The high-frequency micro-vibration components are superimposed on the torque control loop of the motor as a feedforward disturbance compensation term. In a preferred embodiment of the present invention, step S5, which iteratively corrects the electromechanical coupling mathematical model, specifically includes: After the control command is applied, the current dynamic mechanical impedance spectrum is calculated in real time; By comparing the changes in the impedance characteristic vector before and after the control is applied, if the equivalent stiffness coefficient falls back to within the threshold of the normal sliding range and the damping fluctuation tends to be stable, then the jamming is determined to be successfully eliminated. If the impedance eigenvector does not show the expected improvement, the backtracking mechanism is triggered, and a correction signal is fed back to step S4 to readjust the optimization parameters; Meanwhile, the actual response data is used as training samples to correct and update the parameters in the electromechanical coupling mathematical model online.

[0014] This invention addresses the shortcomings of the prior art and has the following beneficial effects: This invention actively applies high-frequency perturbation excitation signals to the drive motor and analyzes the synchronously acquired multi-dimensional state signals based on an electromechanical coupling mathematical model, thereby forcing the transmission mechanism to generate controlled micro-amplitude vibrations. This process reveals the hidden changes in stiffness and damping of the microscopic contact interface, and then extracts the impedance feature vector, achieving highly sensitive online identification of the microscopic friction state. Compared with the passive response mechanism in existing technologies that only rely on macroscopic current or speed thresholds, this solution effectively overcomes the problem that high-frequency weak stick-slip precursor signals are easily masked by background noise and difficult to capture, significantly improving the accuracy and timeliness of sensing the underlying nonlinear friction evolution process in precision assembly lines.

[0015] This invention, based on the system's impedance characteristic vector and combined with a nonlinear dynamic model, predicts the continuous evolution trajectory of system resistance and dynamically determines the jamming risk level by comprehensively calculating the cumulative resistance energy and the probability of exceeding limits. This method transforms discrete system-level characteristic data into a forward-looking projection of future resistance dynamic evolution, thus achieving a mapping from data to trends. Compared to existing technologies that passively trigger alarms only after a severe system lockup or macroscopic resistance mutation, this solution substantially moves the risk prevention checkpoint forward to the microscopic jamming initiation stage. This allows the control system to obtain intervention time windows in advance, avoiding forced emergency shutdowns due to delayed state perception, and laying a technical foundation for flexible control without interrupting the production process.

[0016] This invention constructs a multi-objective optimization function aimed at minimizing jamming probability and cycle time loss, and uses real-time identified impedance parameters as hard constraints to generate a composite control trajectory. Simultaneously, a frequency-domain decoupling strategy is employed to independently map the fundamental wave and micro-vibration components to a dual-loop collaborative drive. This deep electromechanical control integration design effectively prevents control commands from inducing harmful structural resonances and ensures interference-free synchronous execution of macroscopically stable pin feeding and microscopically high-frequency adhesion breaking actions. Compared to existing open-loop strategies that use fixed waveforms for blind jittering or crude reversal, this invention achieves highly adaptive active defense tailored to the instantaneous mechanical characteristics of the system, significantly reducing mechanical losses in the electromechanical system and the production line malfunction rate while efficiently breaking microscopic adhesion points. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. Figure 1 This is a flowchart of the steps of a preferred embodiment of the present invention; Figure 2 This is a block diagram of the hardware and communication architecture of the anti-jamming chain-type automatic nail feeding industrial control system according to a preferred embodiment of the present invention; Figure 3 This is a schematic diagram of the signal acquisition and electromechanical coupling principle of the chain-type automatic nail feeding system according to a preferred embodiment of the present invention. Detailed Implementation

[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein. Therefore, the scope of protection of the invention is not limited to the specific embodiments disclosed below.

[0020] Application Overview: This invention relates to a precision assembly production line, which has high requirements for the continuity of production cycle and assembly accuracy. The nail feeding process, as a highly nonlinear, time-varying electromechanical coupling system, involves the friction state between the chain and fasteners evolving in real time with changes in material batches, ambient temperature and humidity, and the degree of mechanical wear. While existing technologies have introduced pre-jamming detection and deceleration mechanisms, their principles are limited to passive responses to external macroscopic physical thresholds, failing to address the microscopic mechanical essence of jamming. Existing solutions typically treat the nail feeding system as a black box model, focusing only on the apparent relationship between input and output, while ignoring the evolution of dynamic mechanical impedance during chain drive. Control logic based on static thresholds and fixed rules lacks the ability to identify high-frequency, weak stick-slip oscillation signals, leading to frequent malfunctions due to oversensitivity, resulting in production efficiency losses, or failing to provide precise intervention in the early stages of jamming due to lag.

[0021] The fundamental reason why existing technologies struggle to overcome these limitations lies in their methodology, which has long been constrained by traditional discrete event control theory, tending to abstract continuous physical processes into discrete state transition logic. For the specific controlled object of a chain-driven nail feeder, conventional approaches typically rely on increasing the number of sensors to expand the sensing dimension, but fail to construct a mapping model between sensor data and the system's internal physical parameters. If threshold-based discrimination methods or simple multi-sensor data fusion algorithms are directly adopted, the lack of decoupling analysis of the underlying electromechanical coupling dynamics makes it difficult for the system to extract key features characterizing the dynamic state of the microscopic contact interface from the mixed noise. This results in the controller only being able to make response decisions based on macroscopic performance with significant lag, unable to proactively predict the dynamic evolution trajectory of the system state. Furthermore, conventional control strategies often use preset fixed excitation waveforms to perform the de-slip operation. This open-loop excitation ignores the inherent frequency and damping characteristics of the controlled object's real-time evolution, making it difficult to induce structural resonance to achieve the de-slip effect. It may also exacerbate mechanical wear or cause plastic deformation of the nail body due to excitation frequency mismatch, leading to a cycle of sensing lag, misjudgment, and ineffective intervention under complex operating conditions.

[0022] This invention abandons the traditional passive response mode based on rule matching and proposes an active defense strategy based on online identification of dynamic impedance spectrum and nonlinear model predictive control. This strategy no longer relies on a single physical quantity threshold, but instead injects perturbation excitation signals into the system and analyzes the equivalent dynamic mechanical impedance spectrum of the chain drive mechanism in real time, thereby accurately capturing the early characteristics of stick-slip effect at the microscopic level. Based on this, the system constructs an electromechanical coupling predictive model including real-time stiffness, damping, and inertial parameters. Using a nonlinear model predictive control algorithm, a multi-objective optimization function is constructed with the goal of minimizing jamming probability and cycle time loss, and the function is solved in real time to generate a nonlinear velocity-torque composite trajectory adapted to the current system state.

[0023] Exemplary method: like Figure 1 As shown, an industrial control method for an anti-jamming chain-type automatic nail feeding system includes the following steps: S1. During the operation of the chain-type automatic nail feeding system, a high-frequency micro-perturbation excitation signal is applied, and multi-dimensional status signals under the system operation status are acquired, including the real-time current and speed of the drive motor, and the vibration acceleration of the chain-driven mechanism. S2. Based on the electromechanical coupling mathematical model, the multidimensional state signal is analyzed to identify the dynamic mechanical impedance spectrum of the contact interface between the chain and the nail, and the impedance feature vector characterizing the evolution trend of the micro-stick-slip effect is extracted from it. S3. Predict the resistance evolution trajectory of the nail supply system based on the impedance characteristic vector, and then determine the current jamming risk level; S4. When it is determined that there is a risk of jamming, a nonlinear velocity-torque composite control trajectory adapted to the current physical characteristics of the system is generated based on a multi-objective optimization function. S5. Based on the nonlinear speed-torque composite control trajectory drive system, perform adaptive anti-jamming actions, and iteratively correct the electromechanical coupling mathematical model according to the actual system response.

[0024] During the operation cycle of a chain-driven automatic nail feeding system, the dynamic interaction between the drive unit and the transmission mechanism exhibits significant nonlinear dynamic characteristics. This interaction is not only regulated by preset electrical commands but also depends on the instantaneous states of parameters such as chain pitch, nail geometric tolerances, and contact surface friction coefficients. Traditional monitoring methods are typically limited to static threshold comparisons of macroscopic current or average rotational speed, making it difficult to effectively capture the dynamic drift of microscopic mechanical parameters within the transmission chain. When the system is in high-speed reciprocating motion or under variable load conditions, minute fluctuations in mechanical impedance are easily masked by background noise, preventing the control unit from obtaining effective state characterization in the early stages of jamming, thus hindering subsequent intervention measures from receiving accurate data support.

[0025] To overcome the limitations of monitoring a single physical quantity, it is necessary to introduce a specific excitation mechanism on top of the system's conventional drive signal to stimulate and reveal the system's implicit dynamic characteristics. By superimposing a high-frequency perturbation signal of a specific frequency band into the drive loop, the conveyor belt mechanism can be forced to produce micro-amplitude forced vibrations. This vibration response contains rich information about the stiffness, damping, and inertial characteristics of the contact interface. At this point, the system no longer behaves as a passive execution black box, but transforms into a dynamic object that can be detected in real time. Subtle changes in its internal friction state will directly modulate the transmission characteristics of the perturbation signal, forming identifiable feature maps in the current, rotational speed, and vibration acceleration signals at the output end.

[0026] Based on the above mechanism, acquiring comprehensive state data containing perturbation response information becomes the primary prerequisite for achieving high-precision anti-jamming control. This process requires the simultaneous acquisition of multi-dimensional electrical and mechanical parameters of the drive motor under composite excitation, constructing a dataset that reflects the instantaneous dynamic behavior of the system. This data not only records the macroscopic motion trajectory of the system but also encodes the stick-slip evolution law of the microscopic contact surface, providing the necessary raw input for subsequent establishment of electromechanical coupling models, identification of dynamic mechanical impedance spectra, and prediction of drag evolution trends. This ensures that the generation of control strategies has a solid physical foundation and timeliness.

[0027] like Figure 3 As shown, in step S1, a high-frequency perturbation excitation signal is superimposed onto the control circuit of the drive motor, so that the system is simultaneously in a controlled micro-amplitude excitation state under normal production operation conditions. The high-frequency perturbation excitation signal is an additional electrical command with limited amplitude and controlled frequency superimposed on the basic operating command of the drive motor. The perturbation signal satisfies two basic constraints: its amplitude should be sufficiently small to ensure that it does not interfere with the normal nail feeding cycle and conveying accuracy of the system; its frequency band should cover the frequency range of the system's key modes to fully excite the dynamic response of the contact interface.

[0028] The perturbation signal can be implemented using a sinusoidal sweep frequency signal, a pseudo-random binary sequence, or a superimposed multi-frequency simple harmonic wave. Taking a superimposed multi-frequency simple harmonic wave as an example, the perturbation excitation signal can be expressed as: ,in, The perturbation voltage or current component superimposed on the basic drive command at time t; To determine the number of superimposed harmonic waves, multiple frequency points related to the natural frequency of the chain and the engagement frequency of the nails are usually selected. is the amplitude coefficient of the k-th frequency component, which is dynamically adjusted in real time according to the system load to prevent overexcitation; The instantaneous frequency value of the k-th frequency component needs to be pre-calibrated according to the mechanical structure characteristics of the chain drive mechanism, and usually covers a wide frequency range from low-frequency structural modes to high-frequency contact resonance peaks. The initial phase of the k-th frequency component can be randomly distributed to reduce the peak factor and avoid instantaneous impact. By applying a perturbation excitation signal, the microscopic contact state inside the system is forcibly disturbed, allowing the nonlinear characteristics originally hidden under static friction to be manifested in the response signal.

[0029] Step S1 employs a method of superimposing high-frequency perturbation excitation signals and simultaneously acquiring multi-dimensional signals, solving the problems of difficult feature extraction and lagging fault identification in existing technologies. By actively injecting perturbation signals of a specific spectrum, real-time impedance scanning of the mechanical transmission chain is achieved. When minute burrs, dried oil stains, or geometric deformations appear at the contact surface between the chain and the nail, the equivalent stiffness and damping of the contact interface will undergo slight changes. These changes will significantly modulate the transfer function of the high-frequency perturbation signal, thus forming unique fingerprint characteristics in current harmonics and vibration response. Without this active excitation method, these microscopic features would be completely undetectable in normal operating noise.

[0030] After the excitation signal injection is completed, step S1 performs the acquisition of multi-dimensional state signals. Multi-dimensional state signals are defined as a set of physical quantities synchronously acquired from different physical locations and electrical ports of the system under the combined excitation. This set can comprehensively reflect the dynamic behavior of the system, specifically including the real-time current signal and real-time speed signal of the drive motor, as well as the vibration acceleration signal of the chain conveyor mechanism.

[0031] Real-time current signals reflect the instantaneous fluctuations in the motor's output electromagnetic torque. These fluctuations include both the fundamental frequency component overcoming the macroscopic load and the weak modulation sidebands induced by perturbation excitation, making them key electrical parameters for analyzing load abrupt changes and frictional fluctuations. Real-time speed signals characterize the instantaneous angular velocity changes of the motor rotor; given the inertia and damping characteristics of the transmission system, the high-frequency components in the speed signal can be directly mapped to the dynamic response on the chain side. Vibration acceleration signals are acquired through accelerometers installed on the chain guide rail or key nodes of the transmission mechanism. These signals directly record the forced vibration response of the mechanical structure under perturbation excitation and are the most sensitive representation of changes in contact interface stiffness and damping.

[0032] The three physical quantities mentioned above must be strictly synchronized on the time axis. Any tiny timestamp deviation may cause phase distortion during subsequent electromechanical coupling model solving, thus affecting the identification accuracy of the dynamic mechanical impedance spectrum. After preliminary filtering and denoising, the acquired raw data forms a standardized multidimensional state signal vector, which serves as the sole data input source for all subsequent analysis steps.

[0033] After acquiring the multidimensional state signals, this dataset will be used as input for subsequent signal processing and feature extraction. Based on this set of multidimensional state signals rich in high-frequency dynamic features, subsequent processing needs to further decouple the frequency domain parameters reflecting the physical essence of the contact interface from the mixed time domain waveforms. Step S2: The collected signal is analyzed using the constructed electromechanical coupling mathematical model to identify the dynamic mechanical impedance spectrum of the contact interface between the chain and the nail and to extract the impedance feature vector characterizing the evolution trend of the micro-stick-slip effect.

[0034] The electromechanical coupling mathematical model is a set of differential equations that describes the electromagnetic dynamic characteristics of the drive motor, the mechanical dynamics of the chain belt drive mechanism, and the interaction between the two. This model is not a static parameter table, but rather a structured analytical expression containing parameters to be identified. Its mathematical form is based on a pre-defined physical structure and working principle of the system, but key physical parameters need to be estimated and corrected online based on real-time acquired multi-dimensional state signals.

[0035] The dynamic mechanical impedance spectrum is defined as the set of transfer functions describing the response characteristics of a chain drive mechanism under perturbation excitation in the frequency domain. Specifically, it is reflected in the numerical distribution of equivalent stiffness coefficient, equivalent damping coefficient, and equivalent inertial parameter at different frequencies. The impedance eigenvector is a low-dimensional feature extracted from this impedance spectrum, used to quantitatively characterize the evolution stage and severity of the microscopic stick-slip effect.

[0036] Constructing an electromechanical coupling mathematical model requires physical decoupling and mathematical abstraction of the chain-driven automatic nail feeding system. From the drive side, a permanent magnet synchronous motor or stepper motor generates electromagnetic torque under voltage drive, and this torque has a definite electromagnetic relationship with the motor speed and current. From the load side, the chain transmission mechanism can be equivalently represented as a multi-degree-of-freedom mass-spring-damped system, where the elastic deformation of the chain links, the frictional resistance of the guide rail, and the contact stiffness between the nail and the chain collectively constitute the system's mechanical impedance. The motor rotor and the chain drive pulley are rigidly connected via a coupling or reduction mechanism, forming an electromechanical coupling interface. The electromechanical coupling mathematical model can be described in state-space equation form as follows: ,in, The quadrature-axis current driving the motor is the effective current component that generates electromagnetic torque. This is the control voltage applied to the motor's quadrature axis; For the stator inductance of the motor; Stator resistance; The back electromotive force coefficient; This refers to the mechanical angular velocity of the motor rotor. This is the equivalent moment of inertia of the motor rotor and the driving wheel; This is the motor torque coefficient; This is the load torque referred to the motor shaft end; This refers to the angular displacement of the motor rotor; and These are the angular displacement deviation and angular velocity deviation caused by elastic deformation in the chain drive mechanism, respectively. The equivalent stiffness coefficient varies with position, reflecting the elastic compressive characteristics of the interface between the chain link and the nail. The equivalent damping coefficient, which varies with velocity, characterizes the viscous loss properties of the friction pair. The equivalent moment of inertia of the load converted to the motor shaft end includes the mass distribution of the chain and nails.

[0037] In the electromechanical coupling mathematical model, the load torque The dynamic mechanical impedance spectrum is expressed as the sum of three terms: a position-dependent stiffness term, a velocity-dependent damping term, and an acceleration-dependent inertia term. The stiffness and damping coefficients are not predetermined fixed constants, but rather unknown parameters that evolve in real time with the system's operating state. The instantaneous values ​​of these two parameters and their frequency-dependent variations constitute the core content of the dynamic mechanical impedance spectrum. When the nail and chain are under normal lubrication, the stiffness coefficient exhibits linear elastic characteristics, while the damping coefficient remains at a low level. When a microscopic stick-slip effect occurs, the contact interface exhibits alternating local adhesion and sliding phenomena. At this time, the stiffness coefficient exhibits nonlinear hysteresis characteristics, the damping coefficient increases significantly, and resonance peaks appear in specific frequency bands. Therefore, by identifying the spectral distribution of these two parameters online, early identification of impending jamming can be achieved.

[0038] After constructing the electromechanical coupling mathematical model, the model needs to be solved in real time using the multidimensional state signals acquired in step S1 to identify the dynamic mechanical impedance spectrum at the current moment. This process is essentially an inverse problem of parameter estimation: given a known input voltage signal and measurable output current, rotational speed, and vibration signals, the unknown parameters in the model are solved. and The relationship between frequency and the system's nonlinear characteristics and the presence of noise in the measured signal makes conventional least squares estimation or linear recursive methods difficult to apply. This application employs a joint estimation algorithm based on extended Kalman filtering and recursive least squares to track the changes in state variables and parameters in real time in the time domain. Specifically, the differential equations are first discretized into difference equations to establish the state transition matrix and observation matrix; then, the stiffness coefficients and damping coefficients are expanded into state variables to construct an augmented state vector; finally, the system state and model parameters are simultaneously updated in each sampling period using the prediction-update recursive formula of Kalman filtering.

[0039] After obtaining real-time estimates of the stiffness and damping coefficients, it is necessary to further transform them from the time domain to the frequency domain to construct a dynamic mechanical impedance spectrum. This transformation is achieved by performing a Fast Fourier Transform on the estimated sequence over a time window. Let the current time be t, and take the sequence of stiffness coefficient estimates over the past M sampling periods. After applying a window function to the sequence, a discrete Fourier transform is performed to obtain the complex expressions of the stiffness coefficients at different frequencies: ,in, Let f be the complex stiffness at frequency f, where the real part represents the energy storage stiffness and the imaginary part represents the loss modulus. It is a window function used to suppress spectral leakage; The sampling period is [value]. Similarly, by performing the same transformation on the damping coefficient sequence, the complex damping coefficients can be obtained. Inertial parameters The changes are usually slow and can be obtained through averaging over a long window. and The values ​​at different frequencies are presented as curves, which is the dynamic mechanical impedance spectrum. The peak frequency, peak amplitude, and spectral width of this spectrum are directly related to the stick-slip state of the contact interface: when the stick-slip effect intensifies, the stiffness spectrum will show a significant resonance peak at a certain characteristic frequency, while the loss factor of the damping spectrum near that frequency will increase significantly.

[0040] The process of extracting impedance feature vectors based on dynamic mechanical impedance spectra aims to reduce high-dimensional spectral information into low-dimensional feature vectors to facilitate subsequent classification and predictive analysis. This application selects the following three dimensions of features to construct the impedance feature vector: First, the resonant frequency offset, i.e., the deviation of the current stiffness spectrum peak frequency from the system's reference natural frequency, which reflects the softening or hardening trend of contact stiffness; second, the peak value of the damping loss factor, i.e., the maximum value of the damping spectrum within the resonant frequency band, which quantifies the severity of the friction energy dissipation process; third, the spectral energy distribution entropy, i.e., the Shannon entropy calculated after probability normalization of the stiffness or damping spectrum, which describes the dispersion of energy distribution in different frequency bands. When the stick-slip effect intensifies, energy is more concentrated in the resonant frequency band, leading to a decrease in entropy. These three features together constitute a three-dimensional impedance feature vector, denoted as […]. ,in, This is the resonant frequency offset. The peak value of the damping loss factor. The entropy is the spectral energy distribution.

[0041] Existing technologies typically treat systems as black-box models, focusing only on the external correlation between their inputs and outputs, without establishing mathematical models that accurately reflect the dynamic changes in internal physical parameters. When dealing with weak and nonlinear precursor signals of anomalies such as stick-slip effects, time-domain statistical features, due to their limited sensitivity, are insufficient to effectively detect early anomalies. Usually, the system only triggers an alarm when jamming has already occurred and there are significant changes in current or rotational speed, at which point the optimal intervention window has been missed. Step S2 constructs an electromechanical coupling mathematical model and applies frequency-domain system identification techniques to decouple the dynamic mechanical impedance spectrum from the multidimensional response signal, thereby extracting an impedance feature vector that can quantify the evolution trend of the microscopic stick-slip effect. This lays the foundation for predicting the resistance evolution trajectory and determining the jamming risk level in subsequent steps. Based on this feature vector that reflects the real-time evolution trend of the stick-slip state of the contact surface, the system needs to further extrapolate its dynamic behavior within future time steps to predict potential jamming risks. Specifically, in step S3, the resistance evolution trajectory of the nail supply system is predicted based on the impedance feature vector, thereby determining the current jamming risk level.

[0042] In step S3, based on the impedance feature vector extracted in step S2, a nonlinear state observer is used to predict the resistance evolution trajectory of the nail feeding system within a future time window, and the current jamming risk level is assessed based on the prediction result. The resistance evolution trajectory does not refer to the instantaneous resistance value at a specific moment, but rather a dynamic curve describing the change in equivalent resistance over time or displacement caused by the frictional interaction between the chain and guide rail, and between the nail body and the feed trough during continuous movement of the nail feeding system. This curve contains complete phase transition information of the system from stable sliding to adhesive stagnation. The jamming risk level is a multi-dimensional evaluation index used to characterize the probability of complete system jamming under current operating conditions and predicted trends, and its potential impact on production cycle time. For example, this level can be divided into four levels: safe, attention, early warning, and critical, each corresponding to different control intervention strategies.

[0043] The specific implementation of step S3 relies first on in-depth mining of the time-series data of impedance eigenvectors. Since the impedance eigenvectors generated in step S2 are sampled values ​​at discrete time points, obtaining a continuous drag evolution trajectory requires establishing a dynamic prediction model that can describe the evolution of these eigenvalues ​​over time. This model must fully consider the nonlinear frictional characteristics of the nail supply system, the time-varying nature of its parameters, and the influence of external random disturbances. Specifically, the system takes the impedance eigenvector sequence over several historical periods as input and uses state-space reconstruction technology to map it into a high-dimensional phase space to reveal the deterministic dynamic attractors hidden behind the seemingly chaotic data. Based on this, a kernel-based nonlinear regression algorithm or a long short-term memory neural network architecture is used to construct the mapping relationship between impedance parameters and future drag values. The prediction process not only focuses on the current absolute drag value but also emphasizes analyzing the trends of the drag change rate, acceleration, and higher-order derivatives, because these differential indicators often reflect the loss of system stability earlier than the absolute value.

[0044] The core of predicting the resistance evolution trajectory of a nail feeding system lies in constructing a hybrid prediction equation that integrates the advantages of physical mechanisms and data-driven approaches. This equation transforms the current dynamic stiffness and damping parameters into equivalent frictional forces, and iteratively extrapolates the kinematic state, while also introducing a historical memory term to capture the hysteresis effect of the frictional state. The specific overall representativeness calculation formula is defined as follows: ,in, Represents the equivalent total resistance at the predicted time; The static friction coefficient is used as a benchmark for dynamic correction. The dynamic contact stiffness at the current moment is derived from the real part of the S2 impedance spectrum; It is the amount of micro-elastic deformation, determined by the difference between the theoretical position and the actual displacement; The dynamic contact damping at the current moment originates from the imaginary part of the S2 impedance spectrum; The instantaneous relative velocity; the summation term represents the historical memory effect, where Let be the weight coefficient of the i-th memory term. As a decay factor, it determines the rate at which the influence of past frictional conditions fades; It is a non-linear activation function, and its input is... This is the current impedance eigenvector, used to capture unmodeled higher-order nonlinear friction effects; The zero-mean Gaussian white noise term is used to characterize unpredictable environmental random disturbances. Solving this equation using numerical integration methods yields a continuous curve of drag changing over a future period, i.e., the drag evolution trajectory.

[0045] After acquiring the resistance evolution trajectory, step S3 then implements the logic for determining the jamming risk level. This process abandons the single threshold comparison method and instead adopts a multi-feature fusion criterion based on trajectory morphology. The system first calculates the slope integral value of the predicted trajectory, i.e., the energy accumulation of resistance growth; if this value rises sharply in a short period of time, it indicates that the system is rapidly entering the adhesion region. Secondly, it analyzes the curvature change of the trajectory: drastic curvature fluctuations usually indicate an increase in micro-slip instability, which is a precursor to jamming. In addition, it is necessary to calculate the probability distribution of the predicted resistance peak exceeding the system's maximum driving torque threshold. Combining the above indicators, the system constructs a risk index function and quantifies and maps this function to a preset risk level range.

[0046] For example, when the predicted trajectory is flat and far below the drive limit, the system determines it to be at a safe level; when the trajectory shows a slight upward trend but does not touch the warning line, it is determined to be at a concern level, and the system needs to increase the monitoring frequency; when the trajectory slope increases significantly and the predicted peak approaches the drive limit, it is determined to be at a warning level, and the system prepares to initiate preventive control; when the predicted trajectory clearly shows that it will break through the drive limit in the next cycle and is accompanied by high-frequency oscillation characteristics, it is determined to be at a critical level, and the system immediately triggers an emergency intervention command.

[0047] Step S3 employs a nonlinear dynamic prediction method based on impedance eigenvectors. Given the physical nature of jamming, it is not an instantaneous event but rather an evolutionary process where the microscopic contact surface gradually shifts from sliding to adhesion. This process manifests in the impedance parameters as a nonlinear surge in stiffness and abnormal fluctuations in damping. By utilizing the physically meaningful impedance eigenvectors extracted in step S2, and combining them with a complex prediction model incorporating memory effects and nonlinear modulation, this method can capture early, subtle signs of resistance evolution, thus predicting its future development trend before the macroscopic resistance significantly increases. This method significantly advances the timing of anti-jamming control. By identifying high-risk trajectories at the nascent stage of jamming, the system gains a time window, enabling flexible preventative adjustments without interrupting production, thereby avoiding forced emergency stops or forceful jerking.

[0048] Based on the quantitatively assessed risk level and predicted resistance development trend, the system needs to generate optimal control commands in real time that can dynamically adapt to the current physical characteristics of the contact interface, in order to eliminate potential jamming risks while minimizing interference with the production cycle. Specifically, in step S4, when a jamming risk is determined, a multi-objective optimization model is constructed. The objective function aims to minimize the jamming probability, production cycle loss, and mechanical shock, and the real-time changing equivalent stiffness coefficient and equivalent damping coefficient are incorporated into the model as constraints. Subsequently, a nonlinear velocity-torque composite control trajectory adapted to the current physical characteristics of the system is generated online. This trajectory includes dynamically adjusted oscillation frequency, amplitude, and phase sequences.

[0049] Step S4 generates a specific motor drive command sequence based on the qualitative risk level and quantitative resistance prediction output from step S3. The nonlinear speed-torque composite control trajectory is defined as a continuous curve sequence showing the changes in the motor speed reference value and output torque limit value over time within the future control time domain. This sequence is not a simple step or ramp signal, but a nonlinear waveform dynamically adjusted based on the real-time identified system dynamic stiffness and damping characteristics. Specifically, this trajectory explicitly integrates controlled micro-vibration parameters for breaking the micro-adhesion state, namely, the dynamically adjusted oscillation frequency, amplitude, and phase sequence. The multi-objective optimization function, as a mathematical scalar function, is used to evaluate the quality of the control trajectory. It simultaneously considers multiple conflicting performance indicators such as anti-jamming success rate, minimization of mechanical shock, energy efficiency, and production cycle maintenance. By solving for the extrema of this function, the Pareto optimal solution can be obtained. The equivalent stiffness coefficient and equivalent damping coefficient, as key physical constraints, are directly derived from the impedance spectrum identification results in step S2. They define the mechanical boundary of the system at the current moment to ensure that the generated control commands will not induce harmful resonances in the system or cause structural overload.

[0050] The specific implementation of step S4 begins with real-time modeling and constraint setting of the current system's physical characteristics. The system first reads the dynamic mechanical impedance spectrum parameters calculated in step S2, particularly the instantaneous equivalent stiffness coefficient and equivalent damping coefficient of the contact interface, and uses these as internal time-varying parameters of the control model. Simultaneously, combined with the risk level determined in step S3, the optimized boundary conditions are determined. For example, under a critical level, some smoothness constraints are released to prioritize the success rate of escape, allowing for larger oscillation amplitudes; under a warning level, the rate of change of acceleration is strictly limited to avoid excessive vibration. Subsequently, a state-space model containing the motor dynamics equations, the elastic deformation equations of the transmission chain, and the friction evolution equations is constructed. This model will serve as the predictive model for the optimization solver, used to simulate the system response under different control inputs. In this model, the control input is parameterized as a composite form containing a fundamental velocity component and superimposed micro-vibration components, where the frequency, amplitude, and phase of the micro-vibration are variables to be optimized.

[0051] The solution process for the multi-objective optimization function is the core algorithm logic of step S4. This process adopts a nonlinear model predictive control framework. Within each control cycle, using the current system state as the initial condition, it searches for a set of optimal speed and torque control sequences in the future finite time domain. The optimization objective function is designed as a weighted sum of multiple sub-objectives, including: minimizing the resistance exceedance probability, i.e., ensuring that the predicted resistance is always lower than the available torque of the motor; minimizing the speed tracking error to ensure stable pin feeding cycle; minimizing the jerk (jerkiness) to reduce mechanical shock and noise; and minimizing control energy consumption. To handle the problem that the dimensions of each objective are different and their importance changes dynamically with the risk level, an adaptive weighting mechanism is introduced, adjusting the weight coefficients of each indicator in real time according to the risk index in step S3. The solver uses a sequential quadratic programming algorithm or interior point method, iteratively searching for the control trajectory that minimizes the overall objective function value while satisfying hard constraints such as the maximum motor current, maximum speed, and mechanical strength. Specifically, the optimization process embeds the equivalent stiffness coefficient and equivalent damping coefficient as hard constraints into the system dynamics equations to ensure that the generated oscillation frequency avoids the system’s inherent resonance band, while utilizing damping characteristics to maximize the dissipation efficiency of vibration energy to break the adhesion point.

[0052] The specific formula for calculating overall representativeness is defined as follows: And satisfy and In this formula, The value of the comprehensive cost function to be minimized; the integration interval. to Represents the prediction time domain; The dynamic weighting coefficient for the i-th sub-objective is determined in real time by the risk level; The normalized set of sub-objective terms corresponds to the jamming probability, the squared velocity tracking error, and the squared jerk, respectively; in the constraint equations, For the equivalent quality of the system, These are the real-time equivalent damping coefficient and equivalent stiffness coefficient identified in step S2, respectively. For displacement, The input for the nonlinear speed-torque composite control to be optimized; To control the superimposed oscillation frequency in the trajectory, Based on the current and The calculated system natural frequency is constrained by an inequality that requires the control oscillation frequency to avoid the system natural frequency to prevent resonance amplification. By solving this variational problem with physical constraints, a nonlinear velocity-torque composite control trajectory containing optimal micro-vibration parameters can be obtained.

[0053] After obtaining the optimal solution, step S4 selects only the command from the first moment of the control sequence and applies it to the motor driver, repeating the above optimization process in the next sampling period to construct a rolling optimization mechanism. This mechanism can compensate for model prediction errors and external disturbances in real time, ensuring that the control trajectory is continuously synchronized with the current actual physical state of the system. The generated nonlinear velocity-torque composite trajectory typically presents the following characteristics: torque reserve is increased before detecting a high-resistance region; when passing through the high-resistance region, micro-vibrations with specific frequencies and phases are superimposed to liquefy the friction layer; after passing through, it quickly recovers to a smooth curve under rated operating conditions.

[0054] Existing technologies typically employ constant speed control combined with overcurrent protection strategies to address pin supply resistance fluctuations. Once the current exceeds a threshold, a shutdown or a preset reverse jitter program is triggered. Such open-loop or simple closed-loop control methods neglect the nonlinear frictional characteristics within the system, often leading to overreaction and shutdown during minor jamming, or inability to escape severe jamming due to improper torque injection timing. Some solutions attempt to introduce fuzzy control or PID parameter self-tuning techniques; however, these methods lack forward-looking prediction of future states, only correcting based on current errors. They cannot pre-adjust the motion state before jamming occurs and struggle to simultaneously coordinate multiple conflicting performance indicators, often resulting in compromises. More critically, existing technologies rarely incorporate real-time identified stiffness and damping parameters as hard constraints into the control command generation process, potentially causing control commands to mismatch with the system's current mechanical characteristics, or even triggering resonance.

[0055] The model predictive control method based on a multi-objective optimization function used in step S4 directly substitutes the equivalent stiffness coefficient and equivalent damping coefficient identified in step S2 as constraints into the solution model, thereby achieving accurate simulation and boundary control of the system's dynamic behavior. This allows the generated control trajectory to adapt to the current contact interface characteristics, enabling the system to actively utilize vibration energy to break micro-adhesion, rather than passively waiting for jamming to occur.

[0056] Based on the optimized dynamic control sequence, the system needs to map this sequence to the actual actions of the underlying driver and monitor its execution effect in real time to build a complete closed-loop feedback mechanism. According to the generated nonlinear speed-torque composite control trajectory, the drive chain-type automatic nail feeding system performs adaptive de-carding actions, and iteratively corrects the parameters of the electromechanical coupling mathematical model based on the actual system response after execution, thereby realizing adaptive closed-loop optimization of the control strategy.

[0057] Step S5 converts the control trajectory calculated in step S4 into actual mechanical motion and evaluates the control effect in real time. Step S5 requires implementing a frequency-domain decoupling strategy, decomposing the composite control trajectory generated in step S4 into low-frequency fundamental components and high-frequency micro-vibration components in the frequency domain, and mapping them to the motor's speed loop and torque loop respectively to achieve independent control. The dynamic response of the impedance characteristic vector refers to the real-time feedback data of the mechanical characteristics of the system contact interface after applying the composite control trajectory, i.e., the change in stiffness and damping over time. This data is an objective basis for judging whether the micro-adhesion state has been effectively broken and whether the risk of jamming has been truly eliminated.

[0058] Step S5 is based on frequency domain analysis and signal reconstruction of the trajectory output from step S4. Since the nonlinear speed-torque composite control trajectory generated in step S4 contains a specific frequency, amplitude, and phase oscillation sequence designed to break micro-adhesion, directly inputting it as a single speed command to a traditional servo drive often results in the attenuation of key high-frequency micro-vibration components due to the low-pass characteristics of the drive's built-in filter, leading to adhesion breaking failure. Therefore, this step employs a frequency-domain decoupling strategy. Using digital signal processing technology, the composite trajectory is separated through a set of filters with adjustable cutoff frequencies, thereby extracting the fundamental signal characterizing the macroscopic nail feeding cycle and the high-frequency oscillation signal characterizing the micro-adhesion breaking action. The fundamental signal is fed into the motor's speed control loop as a main speed reference command to ensure the overall continuity of the nail feeding process; the high-frequency oscillation signal, after amplitude scaling and phase calibration, is superimposed on the motor's torque control loop as a feedforward disturbance compensation term, directly acting on the motor stator current, thereby exciting high-frequency micro-vibrations with controllable amplitude at the end of the mechanical transmission chain. This dual-loop collaborative driving mechanism ensures that macroscopic motion and microscopic actions are executed synchronously without interfering with each other.

[0059] While implementing frequency-domain decoupling drive, step S5 simultaneously activates the effect verification logic based on impedance eigenvectors. Within a preset delay window after the control command is applied, the system rapidly acquires motor current and rotor position signals, and reuses the time-frequency analysis algorithm from step S2 to calculate the current dynamic mechanical impedance spectrum in real time. By comparing the changes in impedance eigenvectors before and after control application, with particular attention to the decrease in the equivalent stiffness coefficient and the stability of the equivalent damping coefficient, the jamming elimination effect is quantitatively evaluated. If the equivalent stiffness coefficient is detected to have fallen from the predicted high value to within the threshold of the normal sliding range, and the damping fluctuation tends to be stable, it is determined that the jamming risk has been eliminated, and the system automatically switches back to the conventional constant-speed pin supply mode; if the impedance eigenvector does not show the expected improvement or even deteriorates further, it is determined that the current control trajectory is insufficient or the parameters are mismatched, and the system immediately triggers the backtracking mechanism, feeding back a correction signal to step S4, requiring the weight coefficients in the multi-objective optimization function to be readjusted or the oscillation amplitude constraints to be relaxed, thereby generating a new escape trajectory and executing it again.

[0060] Exemplary system: like Figure 2 As shown, an anti-jamming chain-type automatic nail feeding industrial control system includes: a chain-type nail feeding mechanical execution unit, a servo drive unit, a multi-source sensing unit, a programmable logic controller, and an industrial control computer.

[0061] The mechanical actuator mainly consists of a servo motor, a reduction mechanism, a driving sprocket, a driven sprocket, a chain belt, and a guide rail. Fasteners to be transported are arranged at equal intervals on the chain belt. The servo drive unit uses a high-performance servo driver, capable of real-time switching between speed control and torque control modes, and supports the injection of external analog commands and online adjustment of internal current loop parameters. The multi-source sensing unit includes a photoelectric encoder mounted on the servo motor output shaft, a three-phase current Hall sensor integrated into the servo driver, and piezoelectric accelerometers mounted on key nodes of the chain belt guide rail. The programmable logic controller (PLC), as the core control unit of the system, runs a real-time operating system and has built-in analog-to-digital converter (ADC), digital-to-analog converter (DAC), high-speed counter interface, and industrial Ethernet communication interface. The industrial control computer interacts with the PLC via industrial Ethernet, responsible for executing computationally intensive impedance spectrum identification algorithms and model predictive control optimization solutions, and then sending the calculation results to the PLC in the form of structured control words.

[0062] The Programmable Logic Controller (PLC) integrates a main control program, interrupt service routines, and a parameter storage area. The main control program executes state machine scheduling according to a preset scan cycle, covering initialization, normal pin feeding, impedance identification, risk prediction, and active intervention states. The interrupt service routine responds to encoder and timer interrupts at fixed time intervals longer than the main program's scan cycle. Within the interrupt context, it synchronously acquires and timestamps current, speed, and vibration acceleration signals, ensuring nanosecond-level synchronization accuracy for multi-source data. The parameter storage area is divided into read-only and read-write regions: the read-only region stores inherent system parameters such as motor constants, sprocket transmission ratios, chain link masses, and sensor calibration coefficients; the read-write region is used to dynamically update historical time-series data of the equivalent stiffness coefficient, equivalent damping coefficient, equivalent inertia parameter, and impedance eigenvector acquired through real-time identification.

[0063] A perturbation excitation signal generator is integrated into the programmable logic controller (PLC), and this generator is encapsulated as a function block within the main control program. When the system is in normal pin feeding mode, the perturbation excitation signal generator injects a high-frequency perturbation voltage component into the speed loop input of the servo driver. This perturbation signal is a multi-frequency harmonic superposition sequence, and its mathematical expression is defined by the number of superimposed harmonics, the amplitude coefficient of each frequency component, the instantaneous frequency value, and the initial phase parameter. The number of harmonics is pre-calibrated based on the natural frequency of the chain and the meshing frequency of the fasteners, typically covering a wide frequency range from low-frequency structural modes to high-frequency contact resonance peaks. The amplitude coefficients are normalized, and their values ​​are dynamically and adaptively adjusted according to the real-time load rate of the system to ensure that the amplitude of the perturbation signal is always lower than the preset percentage threshold of the normal operating command amplitude, thereby avoiding interference with the pin feeding cycle and conveying accuracy. The distribution range of the instantaneous frequency value is pre-calibrated based on the mechanical structural characteristics of the chain drive mechanism through frequency sweep experiments; the initial phase adopts a random distribution method to reduce the peak factor and prevent stress damage to the mechanical structure caused by instantaneous impacts. The perturbation excitation signal generator employs direct digital frequency synthesis technology, generating continuously variable frequency components through lookup table and linear interpolation methods. The amplitude, frequency, and phase parameters of each frequency component can be reconstructed in real time by an industrial control computer via a communication interface during system operation to adapt to excitation requirements under different operating conditions.

[0064] This system incorporates a multi-source signal synchronous acquisition module within a programmable logic controller (PLC). This module primarily consists of a hardware triggering unit, a data buffer, and a timestamp management unit. The hardware triggering unit uses the encoder zero-pulse signal from the servo driver as the master clock reference and utilizes a field-programmable gate array (FPGA) to achieve hardware-level synchronous triggering of current, rotational speed, and vibration acceleration signals. The current signal is acquired through the sampling resistor in the servo driver's internal current loop, converted to digital form by an analog-to-digital converter (ADC), and transmitted to the PLC. The rotational speed signal originates from the quadrature pulses output by the photoelectric encoder, processed by a fourth-frequency subdivision circuit, input to a high-speed counter, and converted to instantaneous angular velocity values ​​in the interrupt service routine (ISR). The vibration acceleration signal is acquired through a piezoelectric accelerometer, conditioned by a charge amplifier, and then input to the PLC's analog input channel, undergoing analog-to-digital conversion according to a preset sampling frequency. The data buffer employs a ring-shaped buffer structure to store the timing data of the three physical quantities: current, rotational speed, and acceleration. Each data point is accompanied by a high-precision timestamp generated by the hardware triggering unit. The timestamp management unit establishes a unified time reference based on the microsecond-level system clock of the programmable logic controller, thereby ensuring strict synchronization and alignment of the three physical quantities on the time axis. Any timestamp deviation is limited to within the sampling period to avoid phase distortion during subsequent impedance spectrum identification.

[0065] This system incorporates a dynamic impedance spectrum identification module within the industrial control computer. This module utilizes an electromechanical coupling mathematical model as its core analytical engine, receiving multidimensional state signal vectors from the programmable logic controller (PLC) and calculating the dynamic mechanical impedance spectrum of the chain-to-bolt contact interface in real time using parameter estimation methods. The electromechanical coupling mathematical model employs state-space equations to describe the coupling relationship between the electromagnetic dynamic characteristics of the drive motor and the mechanical dynamics of the chain drive mechanism. The model uses the cross-axis current, rotor mechanical angular velocity, and rotor angular displacement of the drive motor as state variables, the cross-axis control voltage as input variables, and the load torque as output variables. The load torque is expressed in the model as the sum of equivalent stiffness terms related to position, equivalent damping terms related to velocity, and equivalent inertia terms related to acceleration. The equivalent stiffness coefficient and equivalent damping coefficient are defined as key parameters to be identified in real time within the model; these are not fixed constants but rather unknown variables that evolve in real time with the system's operating state. Their instantaneous values ​​and their frequency-dependent variations constitute the core content of the dynamic mechanical impedance spectrum.

[0066] The dynamic impedance spectrum identification module employs a joint estimation algorithm combining extended Kalman filtering and recursive least squares to track changes in state variables and parameters in real time over the time domain. The algorithm first discretizes the differential equations of the electromechanical coupling mathematical model into difference equations, establishing the state transition matrix and observation matrix. Then, it expands the equivalent stiffness coefficient and equivalent damping coefficient into state variables, constructing an augmented state vector. Finally, it updates the system state and model parameters synchronously within each sampling period using the prediction-update recursive formula of the Kalman filter. After obtaining real-time estimates of the equivalent stiffness coefficient and equivalent damping coefficient, the identification module applies a window function to the estimated value sequence from previous preset sampling periods and performs a discrete Fourier transform, thereby obtaining the complex expressions of the equivalent stiffness coefficient and equivalent damping coefficient at different frequencies. These two expressions together constitute the dynamic mechanical impedance spectrum. The peak frequency, peak amplitude, and spectral width characteristics of the impedance spectrum are directly related to the stick-slip state of the contact interface: when the stick-slip effect intensifies, the stiffness spectrum exhibits a significant resonance peak at the characteristic frequency, and the loss factor of the damping spectrum increases significantly near that frequency.

[0067] This system integrates an impedance feature vector extraction module into the industrial control computer. This module constructs a three-dimensional impedance feature vector to characterize the evolution trend of the microscopic stick-slip effect by performing dimensionality reduction processing on the dynamic mechanical impedance spectrum. This feature vector consists of three dimensions: resonant frequency offset, peak value of the damping loss factor, and spectral energy distribution entropy. The resonant frequency offset is defined as the deviation between the current peak frequency of the stiffness spectrum and the system's reference natural frequency, used to quantify the stiffness softening or hardening trend of the contact interface. The peak value of the damping loss factor refers to the maximum value of the damping spectrum within the resonant frequency band, reflecting the severity of the frictional energy dissipation process. The spectral energy distribution entropy is the Shannon entropy calculated by probabilistically normalizing the stiffness or damping spectrum, used to describe the dispersion of vibration energy distribution in different frequency bands; when the stick-slip effect intensifies, energy concentrates in the resonant frequency band, and the entropy value decreases accordingly. The impedance feature vector is updated according to a preset period and stored in the readable and writable area of ​​the parameter storage area, forming a time-series data sequence for subsequent prediction module calls and analysis.

[0068] This system integrates a resistance evolution trajectory prediction module into the industrial control computer. Based on time-series data of impedance eigenvectors, this module employs a nonlinear state observer to predict the resistance evolution trajectory of the nail feeding system within future time windows. The resistance evolution trajectory characterizes the dynamic curve of the equivalent resistance changing over time due to the frictional interactions between the chain and guide rail, and between the nail body and the feed trough during continuous motion of the nail feeding system. The prediction module constructs a hybrid prediction equation that integrates physical mechanisms and data-driven approaches. This equation maps the current dynamic contact stiffness, dynamic contact damping, and instantaneous relative velocity to equivalent frictional forces and iteratively extrapolates based on the kinematic state. Simultaneously, the equation introduces a historical memory term to capture the hysteresis effect of frictional states. The historical memory term, through a weighted summation, characterizes the cumulative impact of frictional states from multiple past sampling periods on the current resistance. The weight coefficients of each memory term follow an exponential decay distribution, and the decay factor determines the rate of fading of the influence of past frictional states. Furthermore, the prediction module introduces a nonlinear activation function, using the current impedance eigenvector as input, to capture unmodeled high-order nonlinear friction effects, and superimposes a zero-mean Gaussian white noise term to characterize random environmental disturbances. Solving this hybrid prediction equation using numerical integration yields a continuous curve showing the drag variation over time within a preset future time window.

[0069] This system integrates a jamming risk level determination module into the industrial control computer. This module calculates multi-dimensional risk indicators based on the resistance evolution trajectory and maps the system state to a preset risk level range. The risk indicators include the integral value of the resistance trajectory slope, the change in trajectory curvature, and the probability distribution of the predicted resistance peak exceeding the drive threshold. The integral value of the resistance trajectory slope characterizes the amount of energy accumulated during the resistance growth process; when this value rises sharply within a preset time window, it indicates that the system is rapidly entering the adhesion region. The change in trajectory curvature reflects the severity of fluctuations in the trajectory shape; severe curvature fluctuations indicate increased micro-slip instability. The probability distribution of the predicted resistance peak exceeding the drive threshold is obtained by comparing the predicted trajectory with the available torque threshold of the motor. The jamming risk level determination module integrates the above indicators into a risk index function and maps the quantified value of this function to four levels: safe, attention, warning, and critical. The safety level corresponds to a predicted trajectory that is flat and well below the drive limit; the attention level corresponds to a trajectory that rises slightly but does not reach the warning line; the warning level corresponds to a trajectory with a significantly increased slope and a predicted peak approaching the drive limit; and the critical level corresponds to a predicted trajectory that clearly indicates it will exceed the drive limit in the next control cycle and is accompanied by high-frequency oscillations. The risk level determination results are stored in the parameter storage area in an enumerated data format and serve as the trigger condition for control strategy switching.

[0070] This system deploys a nonlinear velocity-torque composite trajectory optimization module in the industrial control computer. This module is activated when the jamming risk level reaches a warning or critical state, and solves for the optimal control trajectory based on a nonlinear model predictive control framework. The optimization module first reads the equivalent stiffness coefficient and equivalent damping coefficient output by the dynamic impedance spectrum identification module, using them as time-varying parameters of the control model, and simultaneously sets optimization boundary conditions according to the risk level. Subsequently, the optimization module constructs a state-space model containing the motor dynamics equations, the transmission chain elastic deformation equations, and the friction evolution equations, serving as a predictive model to simulate the system's dynamic response under different control inputs. The control input is parameterized as a composite form of a fundamental velocity component and superimposed micro-vibration components, where the oscillation frequency, amplitude, and phase of the micro-vibration are all variables to be optimized. The optimization objective function is designed as a weighted sum of multiple sub-objectives, including minimizing the drag exceedance probability, minimizing the velocity tracking error, minimizing the jerk, and minimizing control energy consumption. The weight coefficients of each sub-objective adopt an adaptive adjustment mechanism, dynamically adjusting the weight allocation according to the real-time risk index. The optimization solver employs a sequential quadratic programming algorithm, iteratively searching for a control trajectory that minimizes the overall objective function while satisfying hard constraints such as maximum motor current, maximum speed, and mechanical strength. During optimization, equivalent stiffness and equivalent damping coefficients are embedded as hard constraints into the system dynamics equations, forcing the superimposed oscillation frequencies in the control trajectory to avoid the system's natural frequencies calculated based on the current equivalent parameters, thus preventing resonance amplification effects. Furthermore, the optimization module uses a rolling time-domain control strategy, solving for the optimal control sequence in the future preset time domain within each control cycle, but only implementing the command for the first moment of the sequence to the servo driver, repeating the optimization process in the next sampling cycle, thereby compensating for model prediction errors and external disturbances in real time.

[0071] This system integrates a frequency-domain decoupling drive module within a programmable logic controller (PLC). This module receives the nonlinear speed-torque composite control trajectory from the industrial control computer and decomposes it into a low-frequency fundamental component and a high-frequency micro-vibration component in the frequency domain. These components are then mapped to the speed loop and torque loop of the servo driver, respectively. The frequency-domain decoupling drive module employs digital signal processing technology, using a set of filters with adjustable cutoff frequencies to achieve signal separation. The low-frequency fundamental component represents the macroscopic pin feeding cycle and, after digital-to-analog conversion, is fed into the speed control loop of the servo driver as the main speed reference command. The high-frequency micro-vibration component represents the microscopic debonding action and, after amplitude scaling and phase calibration, is superimposed onto the torque control loop of the servo driver. This superimposed component acts directly on the motor stator current as a feedforward disturbance compensation term, thereby exciting high-frequency micro-vibrations with controllable amplitude at the end of the mechanical transmission chain. The frequency-domain decoupling drive module also includes a safety monitoring submodule, which monitors the amplitude and frequency of the high-frequency micro-vibration component in real time to ensure that it operates within preset safety boundaries, thus avoiding cumulative damage to the mechanical structure.

[0072] This system incorporates an adaptive closed-loop correction module within the industrial control computer. Within a preset delay window after the control command is applied, this module utilizes a dynamic impedance spectrum identification module and a time-frequency analysis algorithm to calculate the current dynamic mechanical impedance spectrum in real time. The adaptive closed-loop correction module quantifies the jamming elimination effect by comparing the changes in the impedance eigenvectors before and after control application. Evaluation indicators primarily include the decrease in the equivalent stiffness coefficient and the stability of the equivalent damping coefficient. If the equivalent stiffness coefficient falls from the predicted high value back to within the normal sliding range threshold, and the damping fluctuation tends to stabilize, the jamming risk is considered eliminated. At this point, the adaptive closed-loop correction module sends a state switching command to the programmable logic controller (PLC) to return the system to the normal pinning state. If the impedance eigenvector does not show the expected improvement or further deteriorates, the current control trajectory is deemed insufficient or the parameters are mismatched. The adaptive closed-loop correction module triggers a backtracking mechanism, sending a parameter correction command to the nonlinear velocity-torque composite trajectory optimization module, requesting a readjustment of the weight coefficients in the multi-objective optimization function or a relaxation of the oscillation amplitude constraints. The optimization module then generates a new control trajectory and executes it again.

[0073] This system applies dynamic impedance spectrum online identification technology and nonlinear model predictive control methods to the industrial control of automatic nail feeding systems, enabling proactive detection and precise intervention in the early stages of jamming. The system abandons reliance on single physical quantity thresholds, instead employing high-frequency perturbation excitation and impedance spectrum analysis to capture early characteristics of stick-slip effects at the microscopic scale, significantly improving the timeliness and accuracy of jamming prediction. The system integrates the equivalent stiffness coefficient and equivalent damping coefficient obtained in real-time identification as physical constraints into the control trajectory optimization process, ensuring that the generated control commands precisely match the system's current mechanical characteristics, thereby avoiding the resonance risks and ineffective interventions that may occur in traditional open-loop de-jamming operations. The system adopts a frequency-domain decoupling drive mechanism, separating the macroscopic cycle maintenance and microscopic adhesion breaking tasks in the frequency domain, achieving dual guarantees of production continuity and jamming elimination effectiveness. The system constructs a complete control closed loop covering perception, decision-making, execution, evaluation and correction through an adaptive closed-loop correction module, enabling the control strategy to continuously self-optimize as the physical characteristics of the system evolve, and possessing good adaptability to operating conditions and long-term operational stability.

[0074] Based on the preferred embodiments of the present invention described above, those skilled in the art can make various changes and modifications without departing from the inventive concept. The technical scope of this invention is not limited to the contents of the specification, but must be determined according to the scope of the claims.

Claims

1. An anti-jamming chain-type automatic nail feeding industrial control system, characterized in that, include: Chain-driven fastener feeding mechanism for carrying and feeding fasteners; A servo drive unit, connected to the chain-type nail feeding mechanical actuator, is used to provide driving force and has the function of switching between speed control mode and torque control mode; The multi-source sensing unit includes sensors for acquiring current and speed signals of the drive motor and sensors for acquiring vibration acceleration signals of the chain drive mechanism. The programmable logic controller is communicatively connected to the servo drive unit and the multi-source sensing unit, and has a built-in perturbation excitation signal generator, a multi-source signal synchronous acquisition module and a frequency division domain decoupling drive module. It is used to generate high-frequency perturbation excitation signals, synchronously acquire multi-dimensional state signals, and decompose the composite control trajectory and distribute it to the speed loop and torque loop of the servo drive unit. An industrial control computer is connected to the programmable logic controller and has built-in dynamic impedance spectrum identification module, impedance feature vector extraction module, resistance evolution trajectory prediction module, jamming risk level determination module, nonlinear speed-torque composite trajectory optimization module, and adaptive closed-loop correction module. The industrial control computer is configured to: identify the dynamic mechanical impedance spectrum and extract the impedance feature vector based on the multidimensional state signal using an electromechanical coupling mathematical model; predict the resistance evolution trajectory and determine the jamming risk level based on the impedance feature vector; when a jamming risk is determined, generate a nonlinear speed-torque composite control trajectory based on a multi-objective optimization function and combined with the real-time identified equivalent stiffness coefficient and equivalent damping coefficient; and iteratively correct the electromechanical coupling mathematical model based on the actual system response feedback.

2. The anti-jamming chain-type automatic nail feeding industrial control system according to claim 1, characterized in that: The perturbation excitation signal generator in the programmable logic controller is configured to superimpose an additional electrical instruction with limited amplitude and controlled frequency on the basic operating instruction of the drive motor. The additional electrical instruction is one of a multi-frequency simple harmonic wave superposition sequence, a sinusoidal sweep signal, or a pseudo-random binary sequence, and its frequency band covers a wide range from low-frequency structural modes to high-frequency contact resonance peaks. The multi-source signal synchronous acquisition module is configured to use the encoder zero pulse signal of the servo driver as the main clock reference, and to use the hardware triggering unit to perform nanosecond-level synchronous acquisition of current signal, rotation speed signal and vibration acceleration signal, and to attach a high-precision timestamp to each data point to form a multi-dimensional state signal vector.

3. The anti-jamming chain-type automatic nail feeding industrial control system according to claim 1, characterized in that: The nonlinear speed-torque composite trajectory optimization module is configured to use a nonlinear model predictive control framework to construct a state-space model containing motor dynamics equations, transmission chain elastic deformation equations, and friction evolution equations as a predictive model. The equivalent stiffness coefficient and equivalent damping coefficient identified in real time are embedded as hard constraints into the state-space model. Under the premise of satisfying the motor's maximum current, maximum speed, and mechanical strength constraints, a nonlinear speed-torque composite control trajectory containing dynamically adjusted oscillation frequency, amplitude, and phase sequences is generated by solving a multi-objective optimization function. The frequency-domain decoupling drive module is configured to decompose the nonlinear speed-torque composite control trajectory into a low-frequency fundamental component and a high-frequency micro-vibration component using a filter bank with an adjustable cutoff frequency. The low-frequency fundamental component is mapped to the speed control loop of the servo drive unit, and the high-frequency micro-vibration component is mapped to the torque control loop of the servo drive unit.

4. An anti-jamming chain belt automatic nail feeding industrial control method, as described in any one of claims 1-3, characterized in that, Including the following steps: S1. During the operation of the chain-type automatic nail feeding system, a high-frequency micro-perturbation excitation signal is applied, and multi-dimensional status signals under the system operation status are collected, including the real-time current and speed of the drive motor, and the vibration acceleration of the chain-driven mechanism. S2. Based on the electromechanical coupling mathematical model, the multidimensional state signal is analyzed to identify the dynamic mechanical impedance spectrum of the contact interface between the chain and the nail, and the impedance feature vector characterizing the evolution trend of the micro-stick-slip effect is extracted from it. S3. Predict the resistance evolution trajectory of the nail supply system based on the impedance characteristic vector, and then determine the current jamming risk level; S4. When it is determined that there is a risk of jamming, a nonlinear velocity-torque composite control trajectory adapted to the current physical characteristics of the system is generated based on a multi-objective optimization function. S5. Based on the nonlinear speed-torque composite control trajectory drive system, perform adaptive anti-jamming actions, and iteratively correct the electromechanical coupling mathematical model according to the actual system response.

5. The anti-jamming chain-type automatic nail feeding industrial control method according to claim 4, characterized in that: The process of applying the high-frequency perturbation excitation signal in step S1 includes: Additional electrical commands are superimposed on the basic operating commands of the drive motor. The amplitude of the additional electrical commands is limited to a threshold that does not interfere with the normal nail feeding cycle and conveying accuracy, and its frequency band covers a wide range from low-frequency structural modes to high-frequency contact resonance peaks. While applying the high-frequency perturbation excitation signal, the real-time current and speed of the drive motor and the vibration acceleration of the chain conveyor mechanism are synchronously acquired in time to form a multi-dimensional state signal vector.

6. The anti-jamming chain-type automatic nail feeding industrial control method according to claim 4, characterized in that: The process of extracting the impedance eigenvector in step S2 includes: The stiffness coefficient and damping coefficient can be identified in real time. The stiffness coefficient and damping coefficient within a preset time window are converted in the frequency domain to construct a dynamic mechanical impedance spectrum. The resonant frequency offset, damping loss factor peak value, and spectral energy distribution entropy are selected from the dynamic mechanical impedance spectrum to form the impedance characteristic vector.

7. The anti-jamming chain-type automatic nail feeding industrial control method according to claim 4, characterized in that: Step S3, which involves predicting the resistance evolution trajectory and determining the risk level, includes: The impedance characteristic vector sequence within the historical period is mapped to a high-dimensional phase space to construct the mapping relationship between impedance parameters and future resistance values; A continuous curve of resistance changing with time over a future period can be obtained by numerical integration. The morphological characteristics of the continuous curve and the probability distribution of the predicted resistance peak exceeding the system driving threshold are calculated, and a risk index function is constructed in combination. The risk index function is quantized and mapped to multiple preset risk level ranges.

8. The anti-jamming chain-type automatic nail feeding industrial control method according to claim 4, characterized in that: Step S4, which generates the nonlinear velocity-torque composite control trajectory, includes: The equivalent stiffness coefficient and equivalent damping coefficient identified in real time are read as internal time-varying parameters of the control model, and the optimized boundary conditions are determined in combination with the determined risk level. A state-space model containing the motor dynamics equations, the transmission chain elastic deformation equations, and the friction evolution equations is constructed as a prediction model. Using a model predictive control framework, a set of speed and torque control sequences that minimize the multi-objective optimization function is searched within a future finite time domain. The multi-objective optimization function includes four sub-objectives: minimizing the probability of drag overshoot, minimizing the speed tracking error, minimizing the jerk, and minimizing the control energy consumption. The weight coefficients of each sub-objective are adjusted in real time according to the risk level.

9. The anti-jamming chain-type automatic nail feeding industrial control method according to claim 4, characterized in that: The adaptive de-jamming action performed in step S5 specifically includes: A frequency-domain decoupling strategy is adopted to decompose the nonlinear velocity-torque composite control trajectory into a low-frequency fundamental component and a high-frequency micro-vibration component; The low-frequency fundamental component is sent into the motor's speed control loop as a main speed reference command. The high-frequency micro-vibration component is superimposed on the torque control loop of the motor as a feedforward disturbance compensation term.

10. The anti-jamming chain-type automatic nail feeding industrial control method according to claim 4, characterized in that: Step S5, which iteratively corrects the electromechanical coupling mathematical model, specifically includes: After the control command is applied, the current dynamic mechanical impedance spectrum is calculated in real time; By comparing the changes in the impedance characteristic vector before and after the control is applied, if the equivalent stiffness coefficient falls back to within the threshold of the normal sliding range and the damping fluctuation tends to be stable, then the jamming is determined to be successfully eliminated. If the impedance eigenvector does not show the expected improvement, the backtracking mechanism is triggered to feed back a correction signal to step S4 to readjust the optimization parameters; Meanwhile, the actual response data is used as training samples to correct and update the parameters in the electromechanical coupling mathematical model online.

Images