A rice mill remote control system and method fusing operation data monitoring

Abstract

This application relates to the field of agricultural machinery automation control technology, and discloses a remote control system and method for a rice milling machine that integrates operational data monitoring. The system includes a rice milling machine, data acquisition hardware, an edge controller, and a cloud server. The system synchronously acquires the stator current signal of the variable frequency motor and the measured air pressure signal of the bran discharge duct, extracts the friction state perturbation index, and eliminates the micro-pulsation variance by eliminating bias interference. It uses dynamic phase compensation time to perform time-domain shifting of the micro-pulsation variance sequence, aligning it with the perturbation index sequence in phase. It calculates cross-correlation coefficients for operating condition identification and outputs decoupled control commands to the pressure gate motor or proportional damper, respectively. Simultaneously, a constant flow orifice plate is configured on the system's pneumatic circuit to prevent dust blockage. This invention compensates for the transmission lag of data across physical domains, accurately distinguishes between normal load and resistance changes caused by dust cover, and avoids back pressure loss caused by incorrect pressure gate adjustment.

Description

Technical Field

[0001] This invention relates to the field of agricultural machinery automation control technology, specifically to a remote control system and method for a rice milling machine that integrates operational data monitoring. Background Technology

[0002] In the rice milling process, existing rice milling machine control systems typically use the stator current of the variable frequency motor as a feedback signal to assess operating conditions and adjust the pressure gate opening accordingly. However, in actual production, the increase in motor current stems not only from the normal increase in material milling resistance but also from localized blockages caused by bran powder obscuring the bran discharge screen. Because the control system relies solely on a single electrical characteristic, it cannot distinguish between these two different physical phenomena.

[0003] When the hulling screen holes are blocked, causing an increase in current, the control system often misjudges the situation and incorrectly outputs a command to open the pressure gate to release pressure. This misadjustment lowers the mechanical back pressure in the rice milling chamber, impairs the stability of the hulling environment, directly increasing the broken rice rate in the finished product. Simultaneously, the variable frequency motor will also undergo unnecessary oscillation adjustments. Regarding monitoring the hulling condition, directly installing a pressure transmitter faces the drawback of the pressure-sensing pipeline being easily blocked by high-concentration dust backflow. Furthermore, there is a hydrodynamic transmission delay between changes in the milling load in the machining chamber and changes in air pressure pulsations in the air duct. Existing control methods ignore the spatial-temporal difference in aerodynamic transmission, failing to align the time axis of sensor data across physical domains. This results in insufficient accuracy of multi-source data joint diagnosis and makes it difficult to implement decoupled independent control of the pressure gate and the hulling damper. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides a remote control system and method for rice milling machines that integrates operational data monitoring. This solves the problems of existing control systems, which cannot distinguish between normal load fluctuations and dust blockage due to single current monitoring, leading to misjudgment of operating conditions, resulting in incorrect adjustment of the pressure gate and loss of mechanical back pressure, as well as the problems of cross-physical domain data transmission delays and misalignment, and easy blockage of sensor pipelines.

[0005] To achieve the above objectives, the first aspect of the present invention provides a remote control system for a rice milling machine that integrates operational data monitoring, comprising: The rice milling machine is equipped with a variable frequency motor, a feeding gate, a pressure gate motor, and a proportional damper installed in the bran discharge duct. The data acquisition hardware is divided into an electrical circuit for collecting the stator current of the variable frequency motor and a pneumatic circuit for collecting the air pressure characteristics in the chaff discharge duct. The edge controller establishes signal connections with the physical actuator and data acquisition hardware respectively, and is configured with processing logic for execution feature extraction, dynamic phase alignment and decoupling control determination; The cloud server communicates with the edge controller and is configured with processing logic to perform historical data clustering and iterative evolution of judgment thresholds.

[0006] Preferably, the pneumatic circuit includes an absolute pressure transmitter, a pressure tapping tube, and a constant flow orifice plate; the absolute pressure transmitter is connected to the chaff discharge duct through the pressure tapping tube; the air inlet on the side wall of the pressure tapping tube is connected in series with the constant flow orifice plate and connected to the instrument air source; the instrument air source injects continuous positive pressure airflow into the pressure tapping tube through the constant flow orifice plate to form an airflow barrier that prevents dust backflow.

[0007] Preferably, the electrical circuit includes a current sensor connected to one phase of the stator circuit of the variable frequency motor, and the signal output terminal of the current sensor is connected to the analog input channel of the edge controller.

[0008] Preferably, the edge controller is equipped with a hardware timer and a ring data buffer; the hardware timer is set to trigger the synchronous conversion of dual-channel analog signals to maintain the initial alignment of the signals on the time axis; the depth of the ring data buffer is not less than the number of windows corresponding to the maximum theoretical hysteresis time of the system's physical network, and is used for the advance compensation shift of the micro-pulsation variance sequence.

[0009] Preferably, the cloud server establishes a network communication connection and message parsing with the edge controller through an IoT communication protocol or a long connection mechanism.

[0010] The second aspect of this invention provides a remote control method for a rice milling machine that integrates operational data monitoring, applied to the control system described in the first aspect of this invention, comprising the following steps: The stator current signal of the variable frequency motor stator circuit and the measured air pressure signal in the chaff discharge duct are collected synchronously. Within the set sliding time window, the friction state perturbation index is extracted based on the stator current signal, and the micro-pulsation variance to eliminate constant airflow bias interference is extracted based on the measured air pressure signal. The dynamic phase compensation time under the current working condition is obtained, and a time-domain translation operation is performed on the micro-pulsation variance sequence based on the dynamic phase compensation time so that the micro-pulsation variance sequence is aligned with the friction state perturbation index sequence in phase. Calculate the cross-correlation coefficient between the offset-aligned micro-pulsation variance sequence and the friction state perturbation index sequence; The operating conditions are identified based on the differential state of the friction state perturbation index and the numerical range of the cross-correlation coefficient, and decoupling control commands are output to the pressure gate motor or proportional damper of the rice milling machine based on the identification results.

[0011] Preferably, before synchronously acquiring signals, the control method further includes an initialization and frequency band isolation step: reading the switching carrier frequency set by the variable frequency motor; calculating the upper limit boundary frequency of high-frequency extraction based on the switching carrier frequency, wherein the upper limit boundary frequency of high-frequency extraction is not greater than half of the switching carrier frequency; and using the upper limit boundary frequency of high-frequency extraction as a global limiting parameter for subsequent frequency band decomposition to avoid electromagnetic noise interference from the variable frequency motor.

[0012] Preferably, before acquiring the dynamic phase compensation time, the control method further includes a system identification and calibration step based on active perturbation: outputting a step-type feed reduction control pulse to the feed gate of the rice milling machine; capturing the electrical domain extreme timestamps showing a frictional load reduction response in the electrical domain and the pneumatic domain extreme timestamps showing a fluid resistance reduction response in the pneumatic domain; calculating the time difference between the pneumatic domain extreme timestamps and the electrical domain extreme timestamps, using this as a reference hysteresis time, and using this reference hysteresis time to perform adaptive calibration updates on the pneumatic hysteresis mapping matrix inside the system. This step dynamically updates the mapping matrix using measured response data, correcting parameter deviations caused by fluid channel wear.

[0013] Preferably, the friction state perturbation index is extracted based on the stator current signal, specifically including: performing discrete wavelet packet transform frequency band reconstruction on the stator current signal within the sliding time window; extracting the low-frequency fundamental energy integral and the high-frequency perturbation energy integral within the effective friction perturbation frequency band; calculating the ratio of the high-frequency perturbation energy integral to the low-frequency fundamental energy integral to construct the friction state perturbation index, and further calculating the first-order difference of the index between adjacent time windows.

[0014] Preferably, the micro-pulsation variance extracted based on the measured air pressure signal to eliminate constant airflow bias interference specifically includes: calculating the arithmetic mean of the measured air pressure signal within the sliding time window; calculating the sum of squares of the differences between the value of each discrete sampling point and the arithmetic mean, extracting the micro-pulsation variance, and removing the positive pressure bias constant introduced by the hardware anti-blocking airflow through the translation invariance property of variance operation.

[0015] Preferably, obtaining the dynamic phase compensation time under the current operating condition specifically includes: extracting the operating frequency of the variable frequency motor spindle and the arithmetic mean of the air pressure at the current moment, and using them as input variables; in the calibrated and updated two-dimensional aerodynamic hysteresis mapping matrix, using a bilinear interpolation algorithm to solve for the corresponding dynamic hysteresis time, and converting it into the corresponding discrete data window offset.

[0016] Preferably, the cross-correlation coefficient between the offset-aligned micro-pulsation variance sequence and the friction state perturbation index sequence is calculated by: extracting a set length of historical aligned data sequence from the buffer, and calculating the cross-correlation coefficient between the two using the Pearson cross-correlation coefficient formula; the numerator of the calculation formula is the sum of the products of the two sequences after subtracting their arithmetic mean, and the denominator is the square root of the product of the variances of the two sequences plus a constant used to prevent division by zero.

[0017] Preferably, the operating condition is identified and decoupled control commands are output based on the identification results. Specifically, this includes: when the first-order difference of the perturbation index is greater than or equal to the positive jump judgment threshold and the cross-correlation coefficient is greater than or equal to the correlation judgment threshold, it is defined as a normal peeling response; when the first-order difference of the perturbation index is greater than or equal to the positive jump judgment threshold and the cross-correlation coefficient is less than the correlation judgment threshold, it is defined as a rice bran masking intervention response; if it is defined as a normal peeling response, a fixed-distance displacement pulse is output to the pressure gate motor to reduce the mechanical back pressure; if it is defined as a rice bran masking intervention response, the control closed loop of the pressure gate motor is cut off to keep it locked, and a full-open command is output to the proportional damper to execute feedforward negative pressure pulse cleaning.

[0018] Preferably, when defined as a normal peeling response, the target displacement of the pressure gate is calculated as follows: extract the absolute value of the difference between the first-order difference of the perturbation index and the positive jump judgment threshold, multiply it by a preset displacement conversion ratio coefficient, and generate the target displacement.

[0019] Preferably, the control method further includes a parameter adaptive evolution step: the edge controller reports historical feature statistics to the cloud server according to the processing cycle; the cloud server uses an unsupervised clustering algorithm to perform cluster analysis on the received data and extracts cluster centers representing normal steady-state processing; based on the cluster centers, an exponentially weighted moving average algorithm is introduced to calculate the positive jump judgment threshold and correlation judgment threshold for the next cycle, and sends them to the buffer register of the edge controller to perform parameter hot replacement when the device is idle or stopped.

[0020] This invention provides a remote control system and method for rice milling machines that integrates operational data monitoring. It has the following beneficial effects: 1. This invention collects stator current and chaff discharge air pressure signals, combines dynamic phase compensation time to perform time-domain translation of air pressure micro-pulsation data, and calculates the cross-correlation coefficient between the data and current micro-perturbation data. This can compensate for the physical delay in the aerodynamic transmission process, achieve the goal of aligning the characteristics of gas and electricity across physical domains on the time axis, and enable the control system to accurately distinguish between normal rolling load increases and resistance changes caused by dust obstruction. This solves the problem that traditional single current monitoring mode is prone to misjudging the working condition.

[0021] 2. This invention outputs independent action commands to the pressure gate motor and proportional damper based on the working condition identification results, which can achieve decoupled control of mechanical back pressure and exhaust air volume. When dust interference is detected, the pressure gate closed loop can be locked and the damper negative pressure cleaning can be triggered, which can effectively avoid the phenomenon of the system erroneously opening the pressure gate to release pressure due to misjudgment of working conditions, maintain the stability of mechanical back pressure in the rice milling chamber, thereby helping to reduce the broken rice rate and reduce the energy consumption of frequent adjustment of the variable frequency motor.

[0022] 3. This invention continuously injects positive pressure airflow into the pressure tapping tube by configuring a constant flow orifice plate on the pneumatic detection circuit, and uses variance calculation in the control logic to extract the micro-pulsation characteristics of the air pressure signal. At the same time, by utilizing the translation invariance property of variance calculation, the constant positive pressure bias interference introduced by the anti-blocking airflow is eliminated at the data processing level, thereby ensuring the accuracy of airflow pulsation characteristic extraction and the long-term reliability of system operation. Attached Figure Description

[0023] Figure 1 This is a system architecture topology diagram of the present invention; Figure 2 This is the method flow of the present invention; Figure 3 This is a flowchart of the frequency band isolation and parameter initialization process of the present invention; Figure 4 This is a flowchart of the active perturbation and reference hysteresis time calibration process of the present invention; Figure 5 This is a flowchart of the multi-source signal synchronous acquisition and feature extraction process of the present invention; Figure 6 This is a flowchart of the dynamic phase compensation and data alignment processing of the present invention; Figure 7 This is the decoupling control and anti-misadjustment decision logic tree of the present invention; Figure 8 This is a flowchart of the cloud-based parameter adaptive evolution and iteration process of the present invention; Figure 9 This is a comparison curve of the dynamic response of mechanical back pressure under dust masking interference according to the present invention. Figure 10 The following is a comparison chart of the core macroscopic performance indicators of the present invention during 72 hours of continuous operation, wherein (a) is a comparison chart of the stability verification of processing quality, and (b) is a comparison chart of the energy consumption performance of the whole machine.

[0024] Among them, 10. Rice milling machine; 11. Variable frequency motor; 12. Feed gate; 13. Pressure gate motor; 14. Proportional damper; 20. Edge controller; 30. Electrical circuit; 31. Current sensor; 40. Pneumatic circuit; 41. Absolute pressure transmitter; 42. Pressure tap; 43. Constant current orifice plate; 50. Bran discharge duct; 60. Cloud server. Detailed Implementation

[0025] The technical solutions in 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.

[0026] See attached document Figure 1 , Figure 1 This is a topology diagram of a control system architecture according to an embodiment of the present invention. The present invention provides a remote control system for a rice milling machine that integrates operational data monitoring, comprising: The rice milling machine 10, the edge controller 20, and the cloud server 60 are included. The rice milling machine 10 is equipped with physical actuators, which include a variable frequency motor 11, a feed gate 12, a pressure gate motor 13, and a proportional damper 14.

[0027] A variable frequency motor 11 is connected to the processing spindle of the rice milling machine 10, providing power for the rotation of the sand roller. A feed gate 12 is installed at the feed inlet of the rice milling machine 10, controlling the flow rate of the material to be processed into the processing chamber. A pressure gate motor 13 is connected to the discharge end of the rice milling machine 10, adjusting the mechanical back pressure inside the processing chamber. A proportional damper 14 is installed inside the bran discharge duct 50, adjusting the airflow rate of the air network system to draw out the fallen bran.

[0028] The edge controller 20 is electrically connected to the variable frequency motor 11, the feed gate 12, the pressure gate motor 13, and the proportional damper 14. The system is also equipped with independent data acquisition hardware, which is divided into electrical circuit 30 and pneumatic circuit 40.

[0029] Electrical circuit 30 includes current sensor 31. Current sensor 31 is connected to one phase of the stator circuit of variable frequency motor 11. The signal output terminal of current sensor 31 is connected to the analog input channel of edge controller 20.

[0030] The pneumatic circuit 40 includes an absolute pressure transmitter 41, a pressure tapping tube 42, and a constant flow orifice plate 43. The absolute pressure transmitter 41 is connected to the chaff discharge duct 50 via the pressure tapping tube 42. The air inlet on the side wall of the pressure tapping tube 42 is connected in series with the constant flow orifice plate 43 and then connected to the instrument air source. The instrument air source injects a continuous, small amount of positive pressure airflow into the pressure tapping tube 42 through the constant flow orifice plate 43, forming an airflow barrier at the pressure tapping port to prevent dust backflow into the chaff discharge duct 50. The electrical signal output terminal of the absolute pressure transmitter 41 is connected to the edge controller 20.

[0031] See attached document Figure 2 , Figure 2This is a flowchart of a control method according to an embodiment of the present invention. The present invention provides a remote control method for a rice milling machine that integrates operational data monitoring, comprising the following steps: S100, the edge controller 20 establishes a connection with the cloud server 60 through a network communication interface to obtain the basic operating parameter set and initial matrix basis of the current processing batch. The edge controller 20 reads the switching carrier frequency set by the variable frequency motor 11 via the industrial bus, and configures the frequency band isolation parameters according to the frequency value to avoid electromagnetic noise interference; S200, at the initial stage of establishing the processing conditions for a new batch of materials, the edge controller 20 outputs a step-type material reduction control pulse to the feed gate 12. The edge controller 20 synchronously monitors the transient response waveforms of the electrical circuit 30 and the pneumatic circuit 40, capturing the timestamps of the electrical and pneumatic transient characteristics reaching their extreme points. The edge controller 20 calculates the difference between these two timestamps and uses this difference to perform adaptive calibration updates on the internally stored pneumatic hysteresis mapping matrix. In the formal processing stage (S300), the edge controller 20 synchronously acquires the stator current signal output by the current sensor 31 and the measured air pressure signal output by the absolute pressure transmitter 41 at a set sampling frequency. Within a set sliding time window, the edge controller 20 performs frequency band decomposition calculation on the stator current signal, extracting the characteristic energy of a specific high-frequency band and the fundamental energy of a low-frequency band to construct a friction state perturbation index. Simultaneously, the edge controller 20 calculates the arithmetic mean of the air pressure signal within the time window and extracts the micro-pulsation variance immune to constant airflow bias interference based on the air pressure signal. S400, the edge controller 20 substitutes the extracted spindle operating frequency and the arithmetic mean of air pressure into the calibrated aerodynamic hysteresis mapping matrix to solve for the dynamic phase compensation time. The edge controller 20 applies the data offset corresponding to the compensation time to the acquired micro-pulsation variance sequence and calculates the cross-correlation coefficient between the offset-aligned micro-pulsation variance sequence and the friction state perturbation index sequence; S500, the edge controller 20 determines the processing status based on the change in the friction state perturbation index and the numerical range of the cross-correlation coefficient. When the judgment result indicates a normal peeling response, the edge controller 20 sends a control command to the pressure gate motor 13 to execute a fixed-distance pressure relief action. When the judgment result indicates a masking effect intervention response, the edge controller 20 cuts off the control authority over the pressure gate motor 13 and simultaneously outputs a momentary fully open analog command to the proportional damper 14 to maintain a specific pulse cleaning time; S600, the edge controller 20 reports the feature statistics and mechanism motion vectors within the calculation cycle to the cloud server 60 according to the set time rhythm. The cloud server 60 performs cluster analysis on the received data, reconstructs the judgment boundary thresholds, and sends them down to the edge controller 20.

[0032] To further clarify the implementation of each technical aspect of the present invention, the following will provide a detailed description of the implementation of each functional module involved above and its internal processing flow.

[0033] See attached document Figure 3 When the system is powered on or a new processing batch is started, the equipment needs to establish a communication link and set anti-interference frequency band parameters to provide a reliable computing benchmark for subsequent multi-source data sensing. The specific process includes: The edge controller 20 establishes a data connection with the cloud server 60 through its internally configured network communication interface, thereby receiving an initialization parameter set from the cloud server 60 for the attributes of the current batch of rice being processed. In this embodiment, the initialization parameter set specifically includes a basic operating frequency, a positive jump judgment threshold, a correlation judgment threshold, and an initial matrix basis. The initial values ​​of the aforementioned positive jump judgment threshold and correlation judgment threshold are typically calculated and generated by the cloud server based on the envelope of operating conditions during the historical processing of similar materials, or, in the absence of historical data, a preset conservative engineering experience value is issued to ensure that the judgment logic in the initial stage of system startup does not make serious misjudgments. Among them, the positive jump judgment threshold is used to quantify the warning boundary of a sudden increase in the internal frictional resistance of the rice milling machine. The initial matrix basis is specifically represented as a two-dimensional aerodynamic hysteresis mapping matrix before system identification and calibration. For the communication handshake and message parsing process of the network communication interface, those skilled in the art can configure the MQTT protocol or TCP / IP long connection mechanism to implement it. The underlying interaction logic of its cross-domain network communication is a well-known technology in the field and will not be described in detail here.

[0034] After receiving the basic parameters, the edge controller 20 accesses the operating status register inside the variable frequency motor 11 drive unit via the industrial bus to read the currently set switching carrier frequency of the variable frequency motor 11. As a preferred method, the industrial bus specifically adopts industrial field communication protocols such as Modbus-RTU or EtherCAT to enable the edge controller 20 to acquire the operating parameters of the lower-level physical execution node. The switching carrier frequency refers to the frequency at which the inverter switching devices of the variable frequency motor 11 perform high-frequency switching actions during pulse width modulation.

[0035] The edge controller 20 performs band isolation calculations based on the acquired switching carrier frequency and dynamically configures the upper limit boundary frequency of high-frequency extraction required for data processing. In the actual operation scenario of a rice milling machine, electrical characteristics reflecting subtle changes in mechanical physical friction are usually concentrated in the mid-to-high frequency range, while the high-frequency switching action generated by the operation of the variable frequency motor 11 continuously injects high-frequency current noise into the electrical circuit 30.

[0036] If the frequency band selected for subsequent feature extraction overlaps with the frequency band where the switching noise is located, or if the noise frequency band is folded due to improper sampling frequency settings, the true physical friction features will be severely masked by this electromagnetic noise. Based on the Nyquist sampling theorem and anti-aliasing criteria, the edge controller 20 determines the upper limit boundary frequency for high-frequency extraction of the discrete wavelet packet transform using the following mathematical relationship: ; In the formula, The upper boundary frequency of the discrete wavelet packet transform is calculated and configured for edge controller 20. The switching carrier frequency set for the variable frequency motor 11 read by the edge controller 20; 0.5 is the anti-aliasing and anti-folding coefficient set based on the Nyquist folding frequency criterion.

[0037] Considering that low-pass filters in actual hardware circuits are difficult to achieve ideal vertical cutoff characteristics, a certain frequency transition band is usually required. Therefore, in practical engineering implementations, the edge controller 20 can introduce a safety margin factor of less than 1. The specific value of the upper limit boundary frequency for high-frequency extraction is calculated as follows: ; In the formula, The value range can be set from 0.8 to 0.9. The edge controller 20 writes the calculated high-frequency extraction upper limit boundary frequency into its internal memory as a global limiting parameter when performing discrete wavelet packet transform frequency band reconstruction, thereby avoiding the main electromagnetic noise interference range caused by frequency conversion drive in the data processing process.

[0038] See attached document Figure 4 During rice milling, there is a physical transmission time difference between the moment the bran is peeled off the rice grain, causing a change in motor load, and the moment the bran is discharged into the bran discharge duct 50, causing a negative pressure fluctuation. This time difference is influenced by spatial distance and the aerodynamic characteristics of the material. Due to the roughness of the actual air network duct and the density differences of different bran varieties, this aerodynamic lag time is difficult to solve accurately using a purely fluid dynamics theoretical model. Therefore, this embodiment introduces active perturbation technology in system identification, using the measured physical response time difference to dynamically calibrate the system.

[0039] When the edge controller 20 determines that a new batch of material has entered the rice milling machine 10 and the operating conditions have reached a preliminary stable state, it outputs a very short-cycle step-reduction control pulse to the feed gate 12. In this embodiment, the stable operating condition can be determined by monitoring the average stator current of the variable frequency motor 11 to ensure that the fluctuation amplitude within a preset time window is less than a preset proportion. This preset proportion is typically configured between 3% and 5% to confirm that the material flow filling has reached a continuous state. The specific implementation of the step-reduction control pulse is as follows: the edge controller 20 controls the feed gate 12 to rapidly reduce its opening and maintain it for a very short time, and then immediately restores it to its original opening.

[0040] As a preferred approach, the holding time of the control pulse is typically set to 100 to 300 milliseconds, and the reduction in opening is set to 10% to 15% of the current steady-state opening. This range of settings can create a clearly identifiable physical state change within the system while avoiding substantial impact on overall processing output and rice quality.

[0041] Edge controller 20 synchronously monitors the transient response waveforms of electrical circuit 30 and pneumatic circuit 40, capturing the timestamps of the electrical and fluid parameters reaching their extreme response points. Specifically, due to the instantaneous reduction in feed rate, the mechanical friction load in the processing chamber decreases accordingly. Edge controller 20 continuously monitors and extracts the electrical characteristic sequence reflecting the mechanical friction state, identifying and recording the exact system time when the sequence reaches its first negative trough. To prevent false triggering due to industrial random noise, edge controller 20 determines a valid negative trough based on the following condition: the attenuation of this extreme point relative to the steady-state mean before the perturbation must exceed a preset noise tolerance threshold.

[0042] Once this condition is met, the system time is defined as the electrical domain extreme timestamp. Similarly, with the reduction of mechanical friction, the concentration of rice bran powder actually stripped and entering the rice bran discharge duct 50 drops instantaneously, causing a sudden change in the fluid resistance and negative pressure characteristics within the pipeline network. The edge controller 20 synchronously monitors the air pressure characteristic sequence reflecting the micro-pulsations of the fluid, identifies and records the system time at which the corresponding effective negative response trough of the sequence is generated, and defines it as the aerodynamic domain extreme timestamp. For the identification and location of signal sequence troughs, those skilled in the art can use extreme point detection algorithms or sliding differential zero-crossing point judgment methods. The underlying peak-finding logic is a well-known technology in this field and will not be elaborated here.

[0043] The edge controller 20 calculates the time difference between two physical domain extrema based on the acquired two timestamps, and uses this difference as the baseline hysteresis time to perform calibration updates. The baseline hysteresis time is calculated according to the following formula: ; In the formula, This is the true baseline hysteresis time under current pipeline conditions and specific bran properties; Timestamps of aerodynamic domain extrema captured by edge controller 20; The electrical domain extreme timestamp captured by the edge controller 20.

[0044] To ensure the stability of the algorithm, the edge controller 20 calculates... A reasonableness check will then be performed. If If the result is negative or exceeds the preset maximum delay limit of the physical pipeline network, the edge controller 20 will discard the perturbation result and keep the original matrix data unchanged to avoid algorithm crash dead zones caused by sensor packet loss or abnormal operating conditions. After obtaining a valid reference hysteresis time, the edge controller 20 extracts the current spindle operating frequency and average air pressure as coordinate indices to locate the corresponding node in the internally stored two-dimensional aerodynamic hysteresis mapping matrix.

[0045] Edge controller 20 utilizes the calculated By overwriting the historical values ​​of this coordinate anchor point, adaptive system identification and calibration based on actual physical processes are completed. This allows phase compensation in subsequent calculations to be performed based on the actual aerodynamic properties of the current batch of materials, thereby avoiding engineering calculation deviations caused by static theoretical models.

[0046] See attached document Figure 5 After completing the initial system identification and calibration, the equipment enters the formal continuous processing stage. In order to perceive the microscopic physical state inside the rice milling machine, the system needs to process sensor signals from both electrical and pneumatic channels and eliminate data bias interference caused by hardware anti-blocking design.

[0047] Edge controller 20 synchronously acquires the stator current signal output by current sensor 31 and the measured air pressure signal output by absolute pressure transmitter 41 at a set sampling frequency. In this embodiment, edge controller 20 triggers analog-to-analog conversion through an internal high-speed hardware timer interrupt mechanism to ensure that the two signals form an aligned discrete data sequence on the time axis.

[0048] To facilitate subsequent cross-correlation alignment calculations of data sequences across physical domains, as a preferred approach, the controller sets the sampling frequency of both signals to the same value, i.e., sets... The controller sets a sliding observation time window, and the number of stator current signal sampling points included in this time window is . The corresponding sampling frequency is The number of measured air pressure signal sampling points included is: The corresponding sampling frequency is and constraints The length of the sliding observation time window is usually configured based on the rotation cycle of the rice milling machine's main shaft, for example, to cover 3 to 5 complete machine operation cycles.

[0049] The edge controller 20 performs discrete wavelet packet transform frequency band reconstruction on the stator current signal within the aforementioned sliding time window, calculating high and low frequency characteristic energies. Based on the aforementioned configured upper limit boundary frequency for high-frequency extraction, the controller extracts a low-frequency reconstruction signal containing the grid fundamental energy, and a high-frequency reconstruction signal located within the effective friction disturbance frequency band. Specifically, the frequency band of the low-frequency reconstruction signal is configured to cover the power supply reference frequency of the variable frequency motor 11, while the lower limit frequency of the high-frequency reconstruction signal is set to a harmonic greater than three times the mechanical operating frequency of the variable frequency motor 11, and its upper limit frequency is limited by the aforementioned calculated upper limit boundary frequency for high-frequency extraction. For the specific decomposition tree structure and filter coefficient selection of the discrete wavelet packet transform, those skilled in the art can configure it based on the selected wavelet basis functions; the underlying algorithm principle is well-known in the field and will not be elaborated here. The controller calculates the integral of the low-frequency fundamental energy within the current window according to the following formula. Integral energy of high-frequency disturbance : ; ; In the formula, For the first The amplitude of the low-frequency reconstructed signal corresponding to each sampling point; For the first The amplitude of the high-frequency reconstructed signal corresponding to each sampling point.

[0050] The edge controller 20 constructs a friction state perturbation index based on the calculated energy integral and extracts its transient rate of change. To eliminate the influence of absolute current base drift caused by small fluctuations in the industrial power grid supply voltage, the friction state perturbation index is defined by the ratio of high-frequency and low-frequency energy. : ; To characterize the dynamic abrupt change trend of the perturbation index in the time dimension, the edge controller 20 calculates its first-order difference approximation derivative. : ; In the formula, This is the index number of the current time window; This is the index number of the previous adjacent time window; This represents the time step forward of the sliding time window. In actual signal processing, to ensure the continuity of feature extraction and avoid missing transient changes, a data overlap rate of 50% to 75% is typically set between adjacent sliding time windows. That is, it is obtained by subtracting the time of the overlapping area from the total window time.

[0051] For the aerodynamic data processing branch, the edge controller 20 calculates the arithmetic mean of the measured air pressure signal within the current time window. : ; In the formula, The data collected by the absolute pressure transmitter 41 The measured air pressure signal values. The signal acquired by the absolute pressure transmitter 41 contains a small amount of anti-blockage positive pressure airflow continuously injected by the constant flow orifice plate 43, which is superimposed with a static bias. The arithmetic mean... It reflects the macroscopic air network resistance level after the current system is superimposed with anti-blocking airflow, and the edge controller 20 retains it as a benchmark reference parameter for subsequent operating condition mapping and matching.

[0052] The edge controller 20 extracts the micro-pulsation variance of immune constant airflow bias disturbance based on the measured air pressure signal and its arithmetic mean. : ; Variance calculations exhibit translation invariance. When calculating the squared difference between each discrete data point and the arithmetic mean, the positive pressure bias caused by the hardware anti-blocking purge airflow is effectively removed as a common constant during subtraction. This allows the edge controller 20 to directly extract the high-frequency micro-pulsation features of the fluid caused by sudden changes in the concentration of rice bran powder transported within the pipeline network from broadband data superimposed with environmental background pressure without adding an additional hardware high-pass filter. This achieves decoupling of hardware anti-blocking and signal purity acquisition at the algorithm level.

[0053] See attached document Figure 6 After acquiring the frictional state perturbation index sequence in the electrical domain and the micro-pulsation variance sequence in the aerodynamic domain, these two characteristic sequences representing the same processing state naturally misalign on the time axis due to the transmission process in physical space. To evaluate the actual peeling and bran removal coupling conditions inside the rice milling machine, the system needs to perform physical time alignment and mathematical calculations on the multidimensional discrete signals.

[0054] Edge controller 20 extracts the current spindle operating frequency of variable frequency motor 11 and simultaneously retrieves the arithmetic mean of air pressure stored in the aforementioned calculation. In this embodiment, the spindle operating frequency directly affects the mechanical stirring intensity and material flow rate within the processing chamber, while the arithmetic mean of air pressure reflects the steady-state airflow resistance level within the current bran discharge network. These two non-linearly changing physical parameters jointly determine the dynamic hysteresis characteristics of the detached bran transport in the pipeline. Edge controller 20 combines the two as input variables for subsequent hysteresis queries.

[0055] Based on the two input variables mentioned above, the edge controller 20 performs coordinate addressing in the two-dimensional pneumatic hysteresis mapping matrix after the aforementioned adaptive calibration update. Considering that when the rice milling machine is running continuously, the real-time operating frequency and average air pressure may not fall exactly on the preset discrete grid points of the matrix, as a preferred method, the controller uses a bilinear interpolation algorithm to solve for the dynamic hysteresis time of the current operating condition.

[0056] To prevent pointer out-of-bounds exceptions caused by input parameters exceeding the matrix domain under abnormal operating conditions, the edge controller 20 introduces boundary limiting logic before performing interpolation operations. This logic forcibly clamps coordinate points exceeding the matrix boundary to the nearest valid boundary value. After obtaining the specific dynamic hysteresis time, the controller converts it into a discrete data window offset according to the following formula. : ; In the formula, The number of dynamic phase compensation windows to be obtained is of data type unsigned integer; This is the rounding function; The dynamic hysteresis time is obtained by bilinear interpolation. This is a fixed time step that advances the sliding time window forward.

[0057] The edge controller 20 calculates the number of dynamic phase compensation windows. The controller performs a time-domain translation and alignment operation on the micro-pulsation variance sequence. The fluid transport response acquired by the pressure transmitter typically lags behind the mechanical friction response captured by the current sensor 31 in physical time. Therefore, the controller performs a forward-compensated translation of the micro-pulsation variance data sequence along the time axis (i.e., towards historical time points) in the memory buffer. A number of data nodes are configured to coincide in phase with the friction state perturbation exponential sequence. To prevent array out-of-bounds errors caused by translation operations, the depth of the ring data buffer of the edge controller is configured to be no less than the number of windows corresponding to the maximum theoretical hysteresis time of the system's physical network. This operation enables the electrical fluctuations and airflow pulsations generated by the same internal rolling event to achieve time-dimensional index alignment in the discrete data array.

[0058] After phase offset compensation is completed, the system uses the Pearson cross-correlation coefficient to quantify the degree of coordination between the two in waveform evolution. The edge controller 20 extracts a segment of the same length from the buffer. The historical aligned data sequence, the truncation length Based on the system's dynamic response characteristics, the configuration is typically set to cover 10 to 20 rotational cycles of the variable frequency motor 11 during continuous operation, balancing computational stability with sensitivity to transient changes. The controller calculates the cross-correlation coefficient between the offset-aligned micro-pulsation variance sequence and the friction state perturbation index sequence according to the following formula. : ; In the formula, To capture the first [item] within the window Each corresponds to a frictional state perturbation index; This is the arithmetic mean of the friction state perturbation index within the intercepted window; To extract the phase-compensated first element within the window The variance of micropulse; This is the arithmetic mean of the variance of micro-pulsations within the intercepted window; A very small constant set internally for the controller, for example, 10. -6 .constant The insertion of [a specific parameter] is used to prevent the division-by-zero operation from crashing the dead zone when two sequences are in a stationary state. Cross-correlation coefficient. The calculation results are distributed in the range of -1 to 1. The magnitude of the values ​​quantitatively characterizes the positive and negative correlation between the change in mechanical peeling resistance and the micro-pulsation characteristics of airflow discharge, providing a mathematical basis for distinguishing between normal processing and dust cover failure.

[0059] See attached document Figure 7 After completing the time alignment and feature extraction of multi-source signals, the system needs to identify the operating conditions based on the calculated associated feature quantities, so as to guide the lower-level actuator to execute the mutually exclusive decoupling control strategy to prevent the erroneous adjustment phenomenon that is prone to occur in traditional single threshold control.

[0060] The edge controller 20 defines the boundary conditions between normal peeling response (state A) and bran powder masking intervention (state B) based on the first-order difference of the perturbation index and the compensation cross-correlation coefficient. In this embodiment, when the frictional load inside the rice mill changes due to normal material fluctuations, the sudden change in the perturbation index in the electrical domain is usually accompanied by corresponding airflow micro-pulsations in the bran exhaust network, and the two show a high time correlation. When the internal screen is largely masked by impurities such as bran powder, mechanical friction will also show an abnormal increase, but due to the obstruction of the exhaust channel, the pneumatic domain cannot generate normal coordinated pulsations, resulting in a decrease in the cross-correlation coefficient between the two. To achieve quantitative identification of the above physical phenomena, the edge controller 20 constructs the following set of decision inequalities: When the following conditions are met: and When this occurs, it is defined as a normal peeling response (state A). When the following conditions are met: and When this occurs, it is defined as a bran powder masking intervention response (state B).

[0061] In the formula, The first difference of the perturbation index calculated for the current time window; The positive jump judgment threshold initialized in the aforementioned steps is used to determine whether the mechanical load of the system has experienced a sudden surge. Its value is usually determined based on the upper envelope bias of historical stable operating data. The Pearson cross-correlation coefficient after offset compensation; As a preferred approach, the correlation threshold is set between 0.65 and 0.80. To ensure the completeness of the control logic and avoid decision dead zones, when... When the edge controller 20 determines that the current system is in a normal steady-state processing or within an allowable fluctuation range, it maintains the current state of the physical actuator unchanged.

[0062] When the system is determined to be in normal peeling response (state A), the edge controller 20 outputs a fixed-distance displacement pulse to the pressure gate motor 13 to smoothly reduce the mechanical back pressure. At this time, the system faces a real increase in material grinding resistance. To prevent rice grains from breaking due to excessive compression, the edge controller 20 dynamically calculates the required mechanical displacement based on the magnitude of the sudden change. Target displacement. Calculated using the following linear mapping relationship: ; In the formula, The preset displacement conversion ratio coefficient is determined by the combined calibration of the transmission ratio of the pressure gate's mechanical structure and the step angle of the motor. The calculated target displacement for opening the pressure gate is obtained. After acquiring the target displacement, the edge controller 20 converts it into a corresponding number of drive pulses and sends them to the driver of the pressure gate motor 13, thereby smoothly reducing the mechanical back pressure in the rice milling chamber and restoring it to a suitable milling working range.

[0063] If the system is determined to be in a masked intervention response (state B), the edge controller 20 triggers a dimension reduction and decoupling control mechanism. In this situation, the sudden increase in frictional resistance is often due to a false overload caused by partial blockage of the sieve holes by bran powder. If the conventional logic is continued to open the pressure gate, it will lead to a severe loss of internal rice milling pressure, resulting in a large amount of unhulled brown rice. Therefore, the edge controller 20 forcibly cuts off the control closed loop of the pressure gate motor 13 at the software execution level, maintaining its current opening locked.

[0064] In synchronous operation, the edge controller 20 sends a step command to the proportional damper 14 on the bran discharge pipeline, forming a feedforward strong negative pressure pulse cleaning. Specifically, the edge controller 20 commands the proportional damper 14 to quickly open to 100% and maintain the set cleaning cycle, which is typically configured to cover 2 to 3 spindle rotation cycles. This is used to forcibly remove bran powder hidden in the screen pores by utilizing the instantaneously increased fluid negative pressure suction force.

[0065] During the cleaning pulse execution, the drastic negative pressure change within the chaff discharge duct 50 is transmitted in real time through the pressure tap 42 in the pneumatic circuit 40 and captured and fed back by the absolute pressure transmitter 41. The edge controller 20 can verify the actual hydraulic penetration effect of the cleaning action based on this feedback signal. After the cleaning cycle ends, the damper opening automatically returns to the operating point before the change, and the system control closed loop is reactivated. In addition, to avoid the potential dead zone risk of equipment damage caused by prolonged blockage, the edge controller 20 is equipped with a retry counter. If state B is triggered continuously more than a preset number (e.g., configured to 3 times), a shutdown audible and visual alarm is directly triggered, and the processing program is terminated.

[0066] See attached document Figure 8 During prolonged continuous processing, the objective wear of the internal mechanical transmission components of the rice milling machine 10, coupled with the gradual change in the roughness of the inner wall of the ventilation duct, can easily lead to a slow drift in the system's baseline physical response characteristics. To maintain the long-term robustness of the control algorithm, the system needs to implement a cross-cycle adaptive evolution and parameter iteration mechanism.

[0067] Within a set specific processing cycle, the edge controller 20 extracts feature statistics and historical motion vectors of the mechanism, encapsulates them into data packets, and reports them to the cloud server 60. In this embodiment, the specific processing cycle is typically configured to coincide with the end of a single day's processing or the completion of a single batch of material processing. The edge controller 20 asynchronously collects data using a background non-real-time task scheduling thread. The data includes the mean of the first-order difference of the perturbation index, the mean of the cross-correlation coefficient after offset compensation, the cumulative motion frequency vectors of the pressure gate motor 13 and the proportional damper 14 within the specific cycle, and the aerodynamic hysteresis time node data after adaptive calibration updates, thereby avoiding resource congestion on the local high-frequency control link of the equipment. In response to the actual engineering situation of unstable network links, the edge controller 20 temporarily stores data packets that are not successfully sent in local non-volatile memory, waiting for the network to recover before resuming the transmission.

[0068] The cloud server 60 receives data packets reported from multiple device nodes and performs cluster analysis. To avoid the bias of a single algorithm in identifying multidimensional data, the cloud server 60 employs the K-means unsupervised clustering algorithm for data processing as a preferred approach. Specifically, the cloud server 60 uses the parsed first-order difference mean and cross-correlation coefficient mean as a two-dimensional input feature vector, and sets the number of cluster centers. Value. Considering that rice milling machines typically exhibit several typical physical characteristics during actual operation, such as steady-state processing, transitional operating conditions, and localized blockages, this... The value is typically configured between 3 and 5.

[0069] The cloud server 60 uses Euclidean distance as a similarity metric to iteratively calculate the distribution characteristics of historical data. The initial point selection and iterative convergence calculation process of the K-means algorithm can be implemented using standard machine learning libraries by those skilled in the art; its underlying computational logic is well-known in the field and will not be elaborated upon here. After clustering, the cloud server 60 needs to extract cluster centers representing normal steady-state processing from multiple result clusters.

[0070] Since equipment operates in normal processing mode for most of the time during continuous industrial production, the cloud server 60 automatically identifies the cluster with the largest total number of samples and the most concentrated distribution of data points using statistical algorithms, and extracts its center coordinates. In business terms, these coordinates represent the actual physical baseline response level under the current mechanical wear condition.

[0071] Based on the reconstructed steady-state cluster center and its distribution radius, the cloud server 60 calculates the decision threshold and aerodynamic mapping matrix basis for the next stage. To prevent drastic jumps in the cluster center caused by occasional environmental disturbances, which could lead to control instability on the device side, the cloud server 60 introduces an exponentially weighted moving average algorithm for smooth parameter updates. A positive jump decision threshold is used. For example, its iterative formula is as follows: ; In the formula, Prepare an update threshold for the calculated next cycle; The historical threshold values ​​that are actually used in the current edge controller 20; The high-frequency disturbance boundary coordinates are obtained by adding a set safety envelope bias to the steady-state cluster centers extracted by this K-means clustering. This bias ensures that there is a sufficient safety margin between the sudden jump judgment boundary and the daily steady-state fluctuations. To ensure smooth coefficient updates and physical inertia during the evolutionary process, The value is usually between 0.10 and 0.25.

[0072] Similarly, cloud server 60 uses this mechanism to synchronously iterate and determine the correlation threshold. And the basis time constants in the aerodynamic hysteresis mapping matrix. To prevent dead zones caused by extreme deviations in algorithm evolution, the cloud server 60 performs hard clipping verification on the calculation results before output, forcibly limiting... The magnitude of a single evolutionary change must not exceed 10% of the historical average.

[0073] The cloud server 60 sends the updated parameter set to the edge controller 20 via a downlink network link. Upon receiving the new parameters, the edge controller 20 does not immediately apply them during the current continuous processing to avoid the risk of drastic erroneous adjustments to the physical mechanism caused by sudden threshold jumps during processing. In this embodiment, the edge controller 20 temporarily stores the new parameters in a buffer register. By continuously monitoring the operating frequency of the variable frequency motor 11, it determines whether the device has entered an idle standby or shutdown state before performing a hot-swap operation of the memory address. This asynchronous closed-loop iterative mechanism can offset the hardware performance degradation during the device's lifecycle, achieving safe evolution of the control logic.

[0074] To enable those skilled in the art to more clearly understand the purpose, technical solution, and advantages of this invention, the present invention will be further described in detail below with reference to specific application embodiments, real experimental test data, and corresponding drawings. It should be noted that the embodiments described herein are only for explaining the present invention and are not intended to limit the scope of protection of the present invention.

[0075] I. Specific Application Examples To more intuitively demonstrate the engineering application process of this system, we take a continuous rice processing line with a daily output of 200 tons as an example, where the processed material is conventional indica rice with a moisture content of 14.5%.

[0076] After the equipment undergoes a cold start and establishes cloud communication, the edge controller 20 reads the currently set switching carrier frequency of the variable frequency motor 11 via the industrial bus. The frequency is 4kHz. To avoid high-frequency chopping noise from the inverter, the controller uses a safety margin factor. The value is 0.8, and the upper boundary frequency for high-frequency extraction of discrete wavelet packet transform is calculated. Hz. This parameter is then written into the underlying digital signal processing register, defining the effective calculation range for subsequent frequency band reconstruction.

[0077] When the initial material continuously fills the processing chamber of the rice milling machine 10 and the average current fluctuation is less than 4%, the edge controller 20 triggers the reference delay calibration program. The controller instructs the feed gate 12 to instantly reduce its opening by 15% and maintain it for 200 milliseconds. Under this perturbation, the electrical circuit 30 captures the electrical domain extreme timestamp characterizing the reduction in frictional load. At 08:15:22.150, the pneumatic circuit 40 captured the extreme timestamp of the aerodynamic domain, which was caused by a sudden drop in the amount of rice bran powder transported, resulting in a decrease in the air network resistance. The time is 08:15:22.680. Based on this, the baseline lag time of the current physical pipeline network is calculated. The value is 530 milliseconds, and the edge controller 20 uses this value to perform an adaptive refresh of the internal aerodynamic hysteresis mapping matrix.

[0078] In the fourth hour of continuous processing, the equipment encountered a batch of raw materials with an abnormally high bran powder content. At this time, the stator current of the variable frequency motor 11 surged, and the edge controller 20 calculated the first-order difference of the perturbation index for the current window. The value reached 0.18, far exceeding the set threshold for positive jump detection. (Set to 0.10). In traditional logic without superimposed aerodynamic monitoring, this feature would be directly judged as mechanical overload, thus misleading the control system to incorrectly open the pressure valve to release pressure. However, in this embodiment, the edge controller 20 retrieves the interpolated dynamic hysteresis time (corrected to 515 milliseconds under the current operating condition) and shifts the extracted airflow micro-pulsation variance sequence forward by the corresponding number of data nodes for phase compensation. The cross-correlation calculation results show that the offset-compensated Pearson cross-correlation coefficient... The correlation coefficient is only 0.32, far below the threshold for determining correlation. (Set to 0.70). This indicates that despite the sharp increase in mechanical resistance, no matching negative pressure pulsation occurred within the bran discharge duct 50, which the system confirmed as a bran powder masking intervention response (state B). The edge controller 20 immediately cut off control of the pressure gate motor 13, maintaining the current mechanical back pressure lock, while simultaneously outputting a step command to the proportional damper 14 to instantly reach 100% full opening. After three consecutive spindle rotation cycles of strong negative pressure feedforward cleaning, the bran powder clogging the screen holes was forcibly sucked away and removed. Subsequently, the data fed back by the absolute pressure transmitter 41 showed that the air network resistance returned to the normal baseline, and the system then released the forced full opening state of the proportional damper 14, automatically returning to the normal closed loop.

[0079] After the daily processing task is completed, the edge controller 20 asynchronously packages and reports the feature statistics for the entire day. The cloud server 60 uses the K-means algorithm to cluster massive historical envelopes, identifying a 5% steady-state shift in the system's baseline response coordinates due to slight wear on the mechanical sanding rollers that day. This is achieved by introducing weighting coefficients. Using an Exponentially Weighted Moving Average (EWMA) algorithm with a value of 0.15, the cloud server smoothly updates the next period's data. Boundary parameters are set and sent to the buffer register of edge controller 20, so that hot replacement can be automatically completed when the production line is shut down for standby at night.

[0080] II. Experimental Verification and Effect Comparison To verify the actual technical effectiveness of the aforementioned multi-source sensing decoupling control scheme, a parallel comparative experiment lasting 72 hours was conducted on two rice milling machines with identical hardware platforms. The control group used traditional single-stator current PID closed-loop feedback control, while the experimental group used the decoupling control system based on combined gas and electricity sensing provided by this invention. Both groups of equipment used the same batch of conventional indica rice raw materials during the experiment.

[0081] The test results are as follows: according to Figure 9 It is evident that the dynamic responses of the two systems differ fundamentally when faced with interference conditions caused by dust blockage leading to localized clogging. The traditional single-variable PID control group (dotted line) misinterpreted a sudden increase in resistance as normal overload, incorrectly executing the operation of opening the pressure gate. This resulted in a severe drop and oscillation of the effective mechanical back pressure in the rice milling chamber, reaching an amplitude of 45%, disrupting the normal hulling and milling environment. In contrast, the experimental group of this invention (solid line) accurately identified state B, quickly cutting off the control closed loop to lock the mechanical back pressure after the interference occurred, and initiating feedforward pneumatic cleaning. The system pressure rapidly recovered to the target steady state after extremely minor fluctuations. Therefore, this waveform comparison fully demonstrates the effectiveness of the dimension reduction and decoupling mechanism of this invention in preventing erroneous adjustments and stabilizing the machining environment.

[0082] according to Figure 10 As can be seen, after 72 hours of continuous real-vehicle operation, the superiority of this invention has been fully verified by macroscopic statistical data. From Figure 10 As can be seen in (a), due to the precise mechanical back pressure locking and cleaning control mentioned above, the experimental group of this invention reduced the average broken rice rate from 4.8% in the traditional control group to 2.5%, thereby improving the yield grade and economic added value of the rice; simultaneously combined with Figure 10 (b) It can be seen that the experimental group, through high-confidence decoupling control, eliminated the repeated blind oscillations of the system, ensuring that the variable frequency motor always operated in the high-efficiency range, reducing the power consumption per ton of material processed by the entire machine from 18.6 kWh / t to 17.1 kWh / t, successfully achieving an overall energy saving benefit of 8.1%. In summary, this invention significantly improves the intelligent fault tolerance capability and overall efficiency of continuous production lines.

[0083] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A remote control system for a rice milling machine that integrates operational data monitoring, characterized in that, include: The rice milling machine includes a variable frequency motor for providing processing power, a feed gate for controlling material flow, a pressure gate motor for adjusting mechanical back pressure, and a proportional damper installed in the bran discharge duct for adjusting the bran discharge air volume. The electrical circuit includes a current sensor connected to the stator circuit of the variable frequency motor to collect mechanical friction load characteristics; The pneumatic circuit includes an absolute pressure transmitter connected to the discharge duct to collect the micro-pulsation characteristics of the fluid; The edge controller establishes electrical connections with the variable frequency motor, the feed gate, the pressure gate motor, the proportional damper, the current sensor, and the absolute pressure transmitter, respectively. The edge controller extracts a friction state perturbation index sequence based on the stator current signal acquired by the current sensor, and extracts a micro-pulsation variance sequence based on the measured air pressure signal acquired by the absolute pressure transmitter. After aligning the micro-pulsation variance sequence and the friction state perturbation index sequence with time phase compensation, the edge controller identifies the specific processing state that causes the load change by calculating the time-series cross-correlation coefficient between the two, and then selects to issue targeted decoupling control commands to the pressure gate motor or the proportional damper based on the specific processing state.

2. The rice milling machine remote control system with integrated operation data monitoring according to claim 1, characterized in that, The pneumatic circuit also includes a pressure tapping tube and a constant flow orifice plate. The absolute pressure transmitter is connected to the chaff discharge duct through a pressure tapping pipe, and the air inlet on the side wall of the pressure tapping pipe is connected in series with the constant flow orifice plate and connected to the instrument air source. Within a set sliding time window, the edge controller calculates the arithmetic mean of the measured air pressure signal and extracts the micro-pulsation variance sequence that eliminates constant positive pressure airflow bias interference by calculating the square of the difference between the value of each measured air pressure signal point and the arithmetic mean.

3. The rice milling machine remote control system with integrated operation data monitoring according to claim 1, characterized in that, Within a set sliding time window, the edge controller performs discrete wavelet packet transform frequency band reconstruction on the stator current signal, and calculates the low-frequency reconstruction signal integral containing the fundamental wave energy and the high-frequency reconstruction signal integral located in the friction disturbance frequency band, respectively. The edge controller calculates the ratio of the high-frequency reconstructed signal integral to the low-frequency reconstructed signal integral to construct the friction state perturbation index, and calculates the difference between the friction state perturbation indices in adjacent sliding time windows to obtain the first-order difference of the perturbation index.

4. The rice milling machine remote control system with integrated operation data monitoring according to claim 3, characterized in that, The edge controller reads the switching carrier frequency set by the variable frequency motor via the industrial bus; The edge controller multiplies the switching carrier frequency with the anti-aliasing and anti-folding coefficient and the safety margin coefficient to calculate the upper limit boundary frequency of high-frequency extraction, and uses the upper limit boundary frequency of high-frequency extraction as the upper limit frequency of the high-frequency reconstructed signal during the discrete wavelet packet transform frequency band reconstruction.

5. The rice milling machine remote control system with integrated operation data monitoring according to claim 1, characterized in that, The edge controller is equipped with a pneumatic hysteresis mapping matrix and executes a reference delay calibration program when the rice milling machine reaches a stable operating state. The edge controller outputs a step-reduction control pulse to the feed gate; The edge controller synchronously monitors the transient response waveforms of the electrical circuit and the pneumatic circuit, and captures the electrical domain extreme timestamp of the stator current signal when it has an effective negative trough, and the aerodynamic domain extreme timestamp of the measured air pressure signal when it has an effective negative trough. The edge controller calculates the difference between the aerodynamic domain extreme timestamp and the electrical domain extreme timestamp as the reference hysteresis time, and uses the reference hysteresis time to perform calibration update on the aerodynamic hysteresis mapping matrix.

6. The rice milling machine remote control system with integrated operation data monitoring according to claim 5, characterized in that, The specific logic for the edge controller to perform the time phase compensation alignment and calculate the timing cross-correlation coefficient is as follows: The edge controller extracts the current spindle operating frequency and the arithmetic mean of the measured air pressure signal as coordinate indices, performs bilinear interpolation calculations in the calibrated and updated aerodynamic hysteresis mapping matrix, solves for the dynamic hysteresis time, and converts it into the number of dynamic phase compensation windows. The edge controller shifts the data node corresponding to the number of dynamic phase compensation windows ahead of the micro-pulsation variance sequence on the time axis and aligns it with the friction state perturbation index sequence, and then calculates the Pearson cross-correlation coefficient between the offset and aligned micro-pulsation variance sequence and the friction state perturbation index sequence.

7. The rice milling machine remote control system with integrated operation data monitoring according to claim 6, characterized in that, The specific logic of the edge controller identifying the specific processing state and issuing targeted decoupling control commands is as follows: When the first difference of the perturbation index is greater than or equal to the positive jump judgment threshold, and the Pearson cross-correlation coefficient is greater than or equal to the correlation judgment threshold, the specific processing state is defined as a normal peeling response, and the edge controller sends a control command to the pressure gate motor to execute the pressure relief action. When the first-order difference of the perturbation index is greater than or equal to the positive jump determination threshold, and the Pearson cross-correlation coefficient is less than the correlation determination threshold, the specific processing state is defined as a bran powder masking intervention response. The edge controller cuts off the control authority of the pressure gate motor to maintain back pressure locking, and outputs a full-open command to the proportional damper to maintain the preset pulse cleaning time.

8. The rice milling machine remote control system with integrated operation data monitoring according to claim 1, characterized in that, It also includes cloud servers; The edge controller establishes a connection with the cloud server through a network communication interface; the edge controller reports feature statistics and mechanism motion vectors to the cloud server according to a set time period, wherein the feature statistics include the mean of the first difference of the perturbation index and the mean of the cross-correlation coefficient.

9. The rice milling machine remote control system with integrated operation data monitoring according to claim 8, characterized in that, The cloud server performs cluster analysis on the received feature statistics and extracts the cluster center coordinates that characterize steady-state processing. The cloud server uses an exponentially weighted moving average algorithm to calculate and generate new positive jump judgment thresholds and correlation judgment thresholds based on the cluster center coordinates and historical operating thresholds, and sends them to the edge controller; the edge controller performs the replacement of threshold parameters when the judgment device is in an idle standby or shutdown state.

10. A remote control method for a rice milling machine integrating operational data monitoring, applied to the remote control system for a rice milling machine integrating operational data monitoring as described in any one of claims 1-9, characterized in that, Includes the following steps: Remote parameter configuration: The edge controller receives a set of operating parameters remotely via a network communication interface; Synchronous acquisition: The edge controller synchronously acquires the stator current signal output by the current sensor in the electrical circuit and the measured air pressure signal output by the absolute pressure transmitter in the pneumatic circuit at a set sampling frequency. Feature extraction steps: The edge controller extracts the friction state perturbation index sequence based on the stator current signal and the micro-pulsation variance sequence based on the measured air pressure signal; Phase alignment: The edge controller performs time phase compensation alignment between the acquired micro-pulsation variance sequence and the friction state perturbation index sequence to eliminate data misalignment caused by multi-physical domains in spatial transmission; State identification: The edge controller calculates the time-series cross-correlation coefficient between the micro-pulse variance sequence after time phase compensation and alignment and the friction state perturbation index sequence to identify the specific processing state that causes changes in mechanical load; Decoupling control: The edge controller executes branch control commands based on the identified specific processing state: when the specific processing state is determined to be a normal peeling response, a control command is sent to the pressure gate motor to execute a pressure relief action; when the specific processing state is determined to be a masking intervention response, the control authority over the pressure gate motor is cut off, and a full-open command is output to the proportional damper to maintain the preset pulse cleaning time.

Images