Turning system

Abstract

The invention relates to the technical field of turning system data processing, in particular to a turning system, relates to the technical field of turning system data processing, and discloses a turning system and a control method thereof. The system comprises a multi-source sensor network deployed in a gearbox, a hub and a transmission chain, temperature, humidity and particulate matter concentration signals are collected through environment self-sensing sensor nodes, and an oil pressure protection threshold value is dynamically adjusted by combining a material thermal expansion coefficient and an environment working condition; the closed-loop feedback module reversely optimizes crystal oscillator compensation parameters through hydraulic response delay data, and interlocking events drive the sand and dust noise feature library to update. According to the method, the problem of interlocking false triggering caused by space-time misalignment of multi-source sensor data is solved, and environment self-adaptive threshold convergence control is realized.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing of a turning gear system, and in particular to a turning gear system. Background Art

[0002] A turning system is an auxiliary drive device used in large rotating machinery. It is primarily used to control the rotor's rotation at a very low, constant speed during unit startup, shutdown, or maintenance. Its core function is to maintain dynamic balance between the rotor and stator and eliminate mechanical stress concentrations caused by gravity or temperature gradients. When the equipment is shut down, the rotor undergoes periodic bending deformation due to gravity. The turning system, driven by a gear transmission mechanism meshing with the main shaft, maintains the rotor at a critical speed of several revolutions per minute. At this point, the deflection vectors created by the rotor's centrifugal force and gravity are dynamically offset, thus preventing plastic deformation of metal components. During hot shutdown of a steam turbine unit, the system's continuous rotation also promotes uniform axial dissipation of residual heat within the cylinder, preventing thermal stress cracking caused by localized overheating. The turning system's torque design must comprehensively consider rotor mass distribution, support bearing friction coefficient, and gear transmission efficiency. Its interlocking protection logic must be aligned with parameters such as lubrication system oil pressure and jacking height to ensure that the driving process does not cause axial play or excessive radial vibration.

[0003] The existing turning system has a technical pain point of failure in multi-source heterogeneous signal fusion at the data processing level: due to the nonlinear time-varying characteristics of parameters such as vibration mode, temperature field distribution and bearing oil film thickness in the dynamic operation of the rotor, the traditional discrete data acquisition architecture is difficult to achieve spatiotemporal alignment of multi-sensor signals, resulting in millisecond-level timestamp deviation between the gear meshing state monitoring data and the lubrication system oil pressure feedback value. When the system is in transient working conditions, the step response curve of the axial displacement sensor and the attenuated oscillation waveform of the torque sensor cannot be effectively associated with the data through the existing Kalman filter, causing the interlocking protection mechanism based on threshold judgment to be falsely triggered or delayed in response in the critical speed range, thereby affecting the control accuracy of the turning device under thermal stress mutation scenarios. Summary of the Invention

[0004] In view of the shortcomings of the existing technology, the present invention provides a cranking system, which solves the problem of false triggering of the interlocking protection mechanism caused by the temporal and spatial inaccuracy of multi-source heterogeneous sensor data in the existing cranking system.

[0005] In order to solve the above technical problems, the specific technical solutions of the present invention are as follows: The present invention provides a turning system, comprising: A multi-source sensor network, deployed at the gearbox, hub, and drive train monitoring points, includes environmental self-sensing sensor nodes and vibration and displacement sensors. The environmental self-sensing sensor nodes output temperature, humidity, and particulate matter concentration detection signals to collect multimodal data on the dynamic operation of the rotor; A clock synchronization compensation module receives temperature and humidity detection signals from the multi-source sensor network, dynamically corrects the clock offset of the sensor nodes through an environmental coupling clock synchronization protocol, generates a time-space synchronization data stream, and outputs it to a data bus; an environmental robustness data fusion module, which receives a spatiotemporally synchronized data stream from the data bus, calls compensation parameters from an environmental feature database stored in a local memory, performs adaptive filtering and noise reduction on the torque sensor signal, performs dielectric constant compensation on the oil film thickness sensor data, and dynamically allocates weight coefficients of multi-source data according to real-time environmental conditions, generating a fused data packet and transmitting it to the decision bus; An interlocking protection decision module obtains a fused data packet from the decision bus, extracts fault characteristic frequencies based on third-order spectral kurtosis analysis, matches the environmental operating mode, dynamically adjusts the gearbox oil pressure protection threshold, and generates control instructions for the hydraulic locking device and the pinion meshing mechanism; The closed-loop feedback module collects the response delay data of the hydraulic locking device and the interlock protection trigger record, feeds the response delay data back to the crystal oscillator frequency compensation algorithm of the clock synchronization compensation module, writes the interlock protection trigger record into the historical operating condition log of the environmental feature database, and optimizes the drive system parameters online.

[0006] Furthermore, in the turning system of the present invention, the environment self-sensing sensor node includes: The temperature detection unit is deployed at the axial measurement point of the gear meshing surface and outputs the temperature detection signal to the clock synchronization compensation module; The humidity detection unit is integrated in the periphery of the optical channel of the laser ranging module and outputs the humidity detection signal to the dynamic weight distribution unit; The particle concentration detection unit is connected to the centrifugal fan control terminal of the air curtain protection layer deployed outside the laser ranging module; The output signal of the particle concentration detection unit triggers the centrifugal fan speed adjustment through the PID regulator of the air curtain control unit, thereby maintaining the air flow velocity of the air curtain protective layer in a negative correlation with the particle concentration.

[0007] Furthermore, in the turning gear system of the present invention, the clock synchronization compensation module performs: Input the signals from the temperature detection unit and the humidity detection unit into the crystal oscillator frequency compensation algorithm to calculate the clock offset correction value of the IEEE1588 protocol; Based on the time base error model of low-temperature crystal oscillator drift and high-temperature thermal noise, the sampling timestamps of vibration sensors in multi-source sensor networks are interpolated and calibrated. The calibrated data stream is transmitted to the data bus via a double-shielded cable, and the grounding impedance of the outer shielding layer of the cable is ≤0.1Ω.

[0008] Furthermore, in the turning system of the present invention, the environmental robustness data fusion module includes: a wavelet threshold noise reduction unit, activating a high-frequency noise stripping operation in response to an output signal of the particle concentration detection unit; Dynamic weight allocation unit: when the detection value of the humidity detection unit is greater than 90%, the data weight coefficient of the magnetic encoder is increased to 0.75, and the data weight coefficient of the laser ranging module is reduced to 0.25; The oil moisture compensation unit receives the dielectric constant detection value of the oil film thickness sensor and calculates the nonlinear correction coefficient in combination with the oil moisture compensation parameter in the environmental characteristic database.

[0009] Furthermore, in the turning system of the present invention, the interlocking protection decision module includes: A third-order spectral kurtosis analysis unit to extract the gear meshing characteristic frequency from the vibration sensor signal of the multi-source sensor network; The sliding mean filter unit activates the sliding window length adaptive adjustment strategy when the particle concentration detection value exceeds 200μg / m³ for 30 seconds; The threshold dynamic adjustment unit calls the material thermal expansion coefficient curve stored in the environmental feature database, combines the gradient data of the temperature detection unit, and dynamically calculates the oil pressure protection threshold compensation amount.

[0010] Furthermore, in the turning system of the present invention, the threshold dynamic adjustment unit performs: Calling the climate zone index table in the environmental characteristics database to match typical environmental patterns according to latitude and longitude coordinates; Calculate the oil pressure threshold compensation based on the gradient data of the medium temperature detection unit and the thermal expansion coefficient curve of the gearbox material; The compensated threshold parameter is input into the temperature control circulation oil circuit of the hydraulic locking device.

[0011] Furthermore, in the turning system of the present invention, the closed-loop feedback module performs: inputting the interlock protection false triggering event data into the sand and dust noise feature library of the environmental feature database, and updating the sliding mean filter window length; Optimizing the slope parameters of the crystal oscillator frequency compensation module based on the response delay data of the hydraulic locking device; The historical rainfall data in the environmental characteristics database is correlated to shorten the triggering interval of the hydrophobic coating self-cleaning cycle.

[0012] Furthermore, in the turning system of the present invention, the dynamic weight allocation unit and the sliding mean filter unit work in coordination: When the particle concentration detection value increases by 100μg / m³, the sliding mean filter window length is shortened by 20%, and the dynamic weight allocation unit increases the weight coefficient of the vibration sensor by 0.3; The output signal of the sliding mean filter unit triggers the operating mode switching instruction of the environmental feature database.

[0013] Furthermore, in the turning system of the present invention, the climate zone index table is linked to the particle concentration detection unit: When the particle concentration detection value exceeds the dust threshold preset in the environmental characteristic database, the gear meshing characteristic frequency template under the desert climate mode is called; According to the typical environmental pattern matched by the longitude and latitude coordinates, the temperature-oil pressure compensation coefficient matrix of the corresponding area is loaded.

[0014] Furthermore, in the turning gear system of the present invention, the optimization of the slope of the crystal oscillator compensation curve includes: Establishing a mapping relationship model between the hydraulic response delay time and the clock synchronization accuracy; Iteratively adjusting the correction coefficient of the crystal oscillator frequency compensation module based on a gradient descent algorithm; The optimized slope parameters are written into the register group of the FPGA programmable logic device.

[0015] Beneficial effects of the present invention: The present invention effectively suppresses the risk of interlock protection false triggering caused by spatiotemporal misalignment of sensor data through a spatiotemporal synchronization compensation mechanism for a multi-source sensor network and an environmentally adaptive data fusion strategy. An environmentally coupled clock synchronization protocol dynamically corrects the effects of temperature and humidity on crystal oscillator drift, generating a spatiotemporal synchronization data stream with millisecond-level precision and eliminating timestamp deviations in multi-source signals. A wavelet threshold noise reduction unit combines dust concentration detection values ​​to remove high-frequency noise, while a dynamic weight allocation mechanism improves data fusion credibility through humidity-related optical error compensation, addressing the spatiotemporal alignment failure problem of traditional discrete data acquisition architectures. The interlock protection decision module extracts the gear meshing characteristic frequency based on third-order spectral kurtosis analysis and dynamically adjusts the oil pressure protection threshold based on the material thermal expansion coefficient and environmental conditions, reducing threshold criterion drift under transient conditions. A closed-loop feedback module reversely optimizes clock synchronization parameters using hydraulic response delay data. Interlock event characteristics drive online updates of the dust noise feature library, forming an environmental parameter-adaptive threshold convergence mechanism, reducing the interlock protection false trigger rate and shortening response delays. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solution of the present invention, the following is a brief introduction to the drawings required for use in the embodiments. Obviously, for ordinary technicians in this field, other drawings can be obtained based on the drawings without paying any creative labor.

[0017] Figure 1 This is a system architecture diagram of a turning system provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0018] In order to make the purpose, technical solutions and advantages of the present invention clearer, the technical solutions of the present invention will be clearly and completely described below in conjunction with the specific embodiments of the present invention and the corresponding drawings. Obviously, the embodiments described are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention. The technical solutions provided by each embodiment of the present invention are described in detail below in conjunction with the drawings. In order to better understand the purpose of the present invention, the present invention is further described in detail below.

[0019] See also Figure 1 The present invention provides a turning system, comprising: A multi-source sensor network, deployed at the gearbox, hub, and drive train monitoring points, includes environmental self-sensing sensor nodes and vibration and displacement sensors. The environmental self-sensing sensor nodes output temperature, humidity, and particulate matter concentration detection signals to collect multimodal data on the dynamic operation of the rotor; A clock synchronization compensation module receives temperature and humidity detection signals from the multi-source sensor network, dynamically corrects the clock offset of the sensor nodes through an environmental coupling clock synchronization protocol, generates a time-space synchronization data stream, and outputs it to a data bus; an environmental robustness data fusion module, which receives a spatiotemporally synchronized data stream from the data bus, calls compensation parameters from an environmental feature database stored in a local memory, performs adaptive filtering and noise reduction on the torque sensor signal, performs dielectric constant compensation on the oil film thickness sensor data, and dynamically allocates weight coefficients of multi-source data according to real-time environmental conditions, generating a fused data packet and transmitting it to the decision bus; An interlocking protection decision module obtains a fused data packet from the decision bus, extracts fault characteristic frequencies based on third-order spectral kurtosis analysis, matches the environmental operating mode, dynamically adjusts the gearbox oil pressure protection threshold, and generates control instructions for the hydraulic locking device and the pinion meshing mechanism; The closed-loop feedback module collects the response delay data of the hydraulic locking device and the interlock protection trigger record, feeds the response delay data back to the crystal oscillator frequency compensation algorithm of the clock synchronization compensation module, writes the interlock protection trigger record into the historical operating condition log of the environmental feature database, and optimizes the drive system parameters online.

[0020] The turning system provided by the present invention collects the rotor operating status and environmental parameters in real time through a multi-source sensor network. The sensor nodes are distributed at key monitoring positions of the gearbox, wheel hub and transmission chain. The environmental self-sensing sensor nodes have built-in temperature, humidity and particulate matter concentration detection units. The temperature detection unit is installed at the axial measurement point of the gear meshing surface to capture the friction heat field distribution of the tooth surface. The humidity detection unit is integrated into the periphery of the optical channel of the laser ranging module to monitor the water vapor condensation state. The particulate matter concentration detection unit is connected to the centrifugal fan control end of the air curtain protective layer. The centrifugal fan speed is adjusted in real time through the PID regulator of the air curtain control unit, so that the air curtain airflow velocity and the particulate matter concentration form a negative feedback control relationship, thereby maintaining the stability of the optical channel transmittance. The temperature and humidity signals of the sensor node are input into the clock synchronization compensation module. This module uses an environmentally coupled clock synchronization protocol and a crystal oscillator frequency compensation algorithm to analyze the impact of temperature and humidity data on the IEEE 1588 protocol clock offset. A time-base error model is established for low-temperature crystal oscillator drift and high-temperature thermal noise. The sampling timestamp of the vibration sensor is corrected using an interpolation calibration algorithm. The calibrated data stream is transmitted to the data bus via a double-shielded cable. The outer shield of the cable adopts a low-impedance grounding design to suppress electromagnetic interference.

[0021] After receiving the spatiotemporally synchronized data stream from the data bus, the environmental robustness data fusion module accesses a pre-installed environmental feature database in local memory. This database contains a climate zone index table and typical environmental operating condition parameters. The torque sensor signal enters an adaptive filtering chain. When the particle concentration exceeds a preset threshold, the wavelet threshold noise reduction unit is activated to remove high-frequency noise components. The oil film thickness sensor data undergoes nonlinear correction using a dielectric constant compensation algorithm, adjusting the dielectric constant compensation coefficient based on the oil moisture content. The dynamic weight allocation unit reads the environmental state matrix in real time. When the humidity exceeds 90%, the weight coefficient of the magnetic encoder data is increased to 0.75, while the weight of the laser ranging data is simultaneously reduced to 0.25 to compensate for the impact of water vapor refraction on optical measurement accuracy. The weight allocation results are then generated into a fused data packet and transmitted to the decision bus.

[0022] The interlocking protection decision module extracts fault characteristic frequencies based on a fused data packet and uses third-order spectral kurtosis analysis to isolate the gear meshing characteristic frequency band from the vibration signal. The extracted spectrum is then matched with the dust impact spectrum in the environmental signature database. Upon identifying persistent dust conditions, a sliding mean filter strategy is activated, dynamically adjusting the filter window length based on the rate of change of particle concentration to suppress transient interference pulses. The dynamic threshold adjustment unit uses the material thermal expansion coefficient curve from the environmental signature database and, combined with the axial temperature gradient data collected by the temperature detection unit, calculates the gearbox oil pressure protection threshold compensation under low-temperature conditions. The compensated threshold parameter is then fed into the temperature-controlled oil circuit of the hydraulic locking device, which uses a PID algorithm to regulate heating power to maintain stable oil viscosity.

[0023] The closed-loop feedback module collects data on the hydraulic locking device's response delay and interlock protection triggering history. This data is fed back via a bus to the clock synchronization compensation module's crystal oscillator frequency compensation algorithm, optimizing the slope parameters of the crystal oscillator compensation curve to improve clock synchronization accuracy. Interlock protection triggering history is written to the environmental signature database's historical operating condition log, driving the dust and noise signature database to update the sliding mean filter parameters online. The system adjusts the hydrophobic coating self-cleaning cycle based on historical rainfall data. During continuous rainy conditions, the trigger interval is shortened to reduce friction losses, forming an environmentally adaptive closed-loop control mechanism.

[0024] Specifically, in the turning system of the present invention, the environment self-sensing sensor node includes: The temperature detection unit is deployed at the axial measurement point of the gear meshing surface and outputs the temperature detection signal to the clock synchronization compensation module; The humidity detection unit is integrated in the periphery of the optical channel of the laser ranging module and outputs the humidity detection signal to the dynamic weight distribution unit; The particle concentration detection unit is connected to the centrifugal fan control terminal of the air curtain protection layer deployed outside the laser ranging module; The output signal of the particle concentration detection unit triggers the centrifugal fan speed adjustment through the PID regulator of the air curtain control unit, thereby maintaining the air flow velocity of the air curtain protective layer in a negative correlation with the particle concentration.

[0025] The environmental self-sensing sensor node described in the present invention forms a closed-loop control through multi-dimensional environmental parameter detection and linkage with the actuator. The temperature detection unit uses an infrared temperature sensor, which is installed at the axial measurement point on the gear meshing surface. The temperature detection unit collects the friction heat field distribution data of the tooth surface through a high-temperature resistant packaging structure. The detection signal is transmitted to the clock synchronization compensation module via a shielded wire to compensate for the impact of temperature gradients on the stability of the crystal oscillator. The humidity detection unit is integrated into the optical channel annular groove of the laser ranging module and uses a capacitive sensor to monitor the condensation state of the optical element surface in real time. The detection signal is input into the dynamic weight distribution unit through a differential amplifier circuit. When the relative humidity exceeds the threshold, the confidence degradation mechanism of the optical measurement data is triggered.

[0026] The particle concentration detection unit, located in the air intake duct ahead of the laser ranging module, uses laser scattering to measure suspended particulate matter concentration in real time. The detection signal is transmitted to the air curtain control unit via a 4-20mA current loop. The air curtain protective layer consists of an annular air curtain driven by a centrifugal fan. A PID controller dynamically calculates the fan's target speed based on the rate of change of particle concentration. A vector-controlled inverter adjusts the three-phase asynchronous motor drive frequency, creating an exponentially negative correlation between air curtain air velocity and particle concentration. In the event of a sudden increase in dust concentration, the air curtain flow rate is automatically increased to maintain optical window cleanliness.

[0027] Temperature, humidity, and particulate matter detection data are encapsulated and transmitted via a heterogeneous bus protocol. The temperature signal is prioritized for input into the temperature drift correction algorithm of the clock synchronization compensation module. The humidity data stream undergoes a CRC check and is written into the data buffer of the dynamic weight allocation unit. The particulate matter concentration signal is fed into the air curtain control unit after ensuring transmission reliability through a hardware watchdog circuit. The sampling period of these three detection signals is automatically adjusted based on the rate of change of environmental parameters. In heavy rain conditions, the humidity detection sampling frequency is increased to 10Hz to capture rapid condensation. In sandstorm conditions, the particulate matter concentration detection data transmission priority is set to the highest level.

[0028] Specifically, in the cranking system of the present invention, the clock synchronization compensation module performs: Input the signals from the temperature detection unit and the humidity detection unit into the crystal oscillator frequency compensation algorithm to calculate the clock offset correction value of the IEEE1588 protocol; Based on the time base error model of low-temperature crystal oscillator drift and high-temperature thermal noise, the sampling timestamps of vibration sensors in multi-source sensor networks are interpolated and calibrated. The calibrated data stream is transmitted to the data bus via a double-shielded cable, and the grounding impedance of the outer shielding layer of the cable is ≤0.1Ω.

[0029] The clock synchronization compensation module of the present invention achieves unified time bases for multi-source sensor data through an environmental parameter coupling mechanism. The analog signal output by the temperature detection unit is converted to a digital value via a 24-bit Σ-Δ analog-to-digital converter. This signal, along with the digital signal from the humidity detection unit, is fed into the preprocessing unit of the crystal oscillator frequency compensation algorithm. This preprocessing unit uses a sliding window mean filter to eliminate pulse interference in the signal. The combined temperature-humidity effect feature vector is extracted and fed into an LSTM neural network to predict the crystal oscillator frequency drift and generate the IEEE 1588 protocol clock offset correction coefficient. A time-base error model for low-temperature crystal oscillator drift and high-temperature thermal noise is constructed based on a Fourier series expansion. A linear regression algorithm is used to fit the frequency domain distribution characteristics of the crystal oscillator phase noise under different temperature gradients. During the vibration sensor sampling timestamp calibration process, a cubic spline interpolation algorithm is used to reconstruct missing time node data. The calibration reference point is dynamically selected within the range of -40°C to 85°C based on the output value of the temperature detection unit.

[0030] The interpolated and calibrated vibration data stream is transmitted to the data bus via a double-shielded cable. The cable structure uses a composite design of an outer copper braided shield and an inner aluminum foil shield. The outer shield forms an equipotential connection with the equipment rack via a low-impedance grounding ring. The grounding impedance value is controlled below 80mΩ to suppress common-mode interference. The data transmission protocol uses an improved RS-485 differential signal transmission mechanism. Each data frame contains a 16-bit cyclic redundancy check code and an 8-bit timestamp identifier. Abnormal data packets that fail the check trigger the automatic retransmission mechanism of the clock synchronization compensation module. The payload data of the vibration signal is preprocessed with a Hamming window function, and the characteristic components of different frequency bands are separated using frequency domain diversity technology. The components are then convolved with the temperature-compensated clock reference signal to generate a time-space aligned synchronous data stream.

[0031] Specifically, in the turning system of the present invention, the environmental robustness data fusion module includes: a wavelet threshold noise reduction unit, activating a high-frequency noise stripping operation in response to an output signal of the particle concentration detection unit; Dynamic weight allocation unit: when the detection value of the humidity detection unit is greater than 90%, the data weight coefficient of the magnetic encoder is increased to 0.75, and the data weight coefficient of the laser ranging module is reduced to 0.25; The oil moisture compensation unit receives the dielectric constant detection value of the oil film thickness sensor and calculates the nonlinear correction coefficient in combination with the oil moisture compensation parameter in the environmental characteristic database.

[0032] The environmental robustness data fusion module described in this invention optimizes data credibility under complex operating conditions through multimodal data processing and dynamic parameter adjustment. The wavelet threshold denoising unit uses a discrete wavelet transform to perform multi-scale decomposition of the torque sensor signal. When the output value of the particle concentration detection unit exceeds a preset threshold, the wavelet basis function selection strategy is dynamically adjusted based on the frequency domain characteristics of the dust impact noise. Adaptive soft threshold processing is performed on the high-frequency subband coefficients to remove transient noise components caused by particle collisions. The denoised torque signal is smoothed using a Butterworth low-pass filter and output to the data buffer of the dynamic weight allocation unit.

[0033] The dynamic weight allocation unit establishes an error correlation model between the magnetic encoder and the laser ranging module. When the humidity detection unit's output exceeds 90%, it calculates the cumulative error compensation for the laser ranging based on the optical refractive index curve. The weight coefficient of the magnetic encoder data is increased to 0.75 to enhance mechanical position detection accuracy, while the weight coefficient of the laser ranging module is reduced to 0.25. The weight allocation parameters are dynamically updated via a sliding window mechanism. The window length is adaptively adjusted based on the rate of change of environmental parameters. In heavy rain conditions, it is shortened to 5 seconds to quickly respond to sudden changes in humidity.

[0034] The oil moisture compensation unit receives the dielectric constant value collected by the oil film thickness sensor, calls the oil moisture content-dielectric constant mapping table stored in the environmental characteristics database, and uses the least squares method to fit the nonlinear correction coefficient. The compensation algorithm incorporates a temperature gradient compensation factor and performs a secondary calibration of the correction coefficient based on the gearbox axial temperature distribution data. The compensated oil film thickness data passes a CRC check and is input into the interlock protection decision module. The oil condition assessment results are simultaneously fed back to the lubrication system control unit, triggering adjustments to the oil circulation pump's start and stop logic.

[0035] The above-mentioned processing units realize cross-module information interaction through the data bus: the clean torque data output by the noise reduction unit and the position data processed by the weight distribution unit are input into the data fusion engine after being aligned in the time domain. The fusion engine uses the weighted Kalman filter algorithm to generate multi-source data fusion results; the oil moisture compensation data is input into the covariance matrix update module of the fusion engine as an auxiliary correction parameter, forming an adaptive data fusion mechanism driven by environmental parameters.

[0036] Specifically, in the turning system of the present invention, the interlocking protection decision module includes: A third-order spectral kurtosis analysis unit to extract the gear meshing characteristic frequency from the vibration sensor signal of the multi-source sensor network; The sliding mean filter unit activates the sliding window length adaptive adjustment strategy when the particle concentration detection value exceeds 200μg / m³ for 30 seconds; The threshold dynamic adjustment unit calls the material thermal expansion coefficient curve stored in the environmental feature database, combines the gradient data of the temperature detection unit, and dynamically calculates the oil pressure protection threshold compensation amount.

[0037] The interlocking protection decision module described in the present invention achieves precise control of the protection threshold through multi-dimensional feature extraction and dynamic compensation of environmental parameters. The third-order spectral kurtosis analysis unit performs a joint time-frequency domain analysis on the vibration sensor signal, decomposing the vibration signal into 32 frequency bands using a complex Morlet wavelet basis function. The spectral kurtosis value of each sub-band signal is calculated and a kurtosis-frequency matrix is ​​constructed. The dominant frequency band containing the gear meshing characteristic frequency is extracted using a peak search algorithm, eliminating pseudo-frequency components caused by environmental vibration noise. The extracted characteristic frequency sequence is then correlated with the gear wear pattern template stored in the environmental feature database to generate a gear health status assessment index.

[0038] The sliding mean filter unit maps particulate matter concentration to signal noise intensity. If the measured particulate matter concentration exceeds 200 μg / m³ for 30 seconds, the filter window length is dynamically contracted based on the concentration gradient. The window length adjustment strategy uses an exponential decay function model, shortening the lower limit of the window length by 15% for every 50 μg / m³ increase in concentration. The adjusted sliding mean filter smoothes the torque signal in the time domain to suppress transient pulse interference caused by dust impact. The filtered output signal is then fed into the threshold judgment module after passing through a phase compensator to eliminate time delay errors.

[0039] The threshold dynamic adjustment unit uses thermal expansion coefficient curves stored by material grade in the environmental characteristic database. Combined with the gearbox axial temperature gradient distribution data collected by the temperature detection unit, a piecewise linear interpolation algorithm is used to calculate the metal component deformation compensation coefficient under low-temperature conditions. A nonlinear temperature-pressure coupling model is used to calculate the oil pressure protection threshold compensation. The relationship between the compensation coefficient and the temperature gradient is fitted using Legendre polynomials. The dynamically compensated threshold parameters are then written into the PID controller register of the hydraulic locking device. The compensation logic triggers heating power adjustment commands for the temperature-controlled oil circuit, converging the oil viscosity control error to within ±2%.

[0040] The above-mentioned processing units form a closed-loop control link: the gear health index output by the spectral kurtosis analysis unit is input into the adaptive parameter adjustment module of the sliding mean filter unit to drive the optimization of the filtering strategy; the clean torque signal after filtering and the compensation parameter output by the threshold dynamic adjustment unit are jointly input into the oil pressure protection decision device to generate the hydraulic actuator control instruction; the execution result of the protection action is fed back to the environmental feature database via the bus, driving the online correction and update of the thermal expansion coefficient curve.

[0041] Specifically, in the turning system of the present invention, the threshold dynamic adjustment unit performs: Calling the climate zone index table in the environmental characteristics database to match typical environmental patterns according to latitude and longitude coordinates; Calculate the oil pressure threshold compensation based on the gradient data of the medium temperature detection unit and the thermal expansion coefficient curve of the gearbox material; The compensated threshold parameter is input into the temperature control circulation oil circuit of the hydraulic locking device.

[0042] The threshold dynamic adjustment unit described in the present invention achieves dynamic optimization of the oil pressure protection threshold through the multidimensional coupling of geographic environmental characteristics and material properties. A climate zoning index table, constructed based on the Köppen climate classification method and stored in distributed nodes within the environmental characteristics database, maps latitude and longitude coordinates to typical environmental patterns. The index table uses the GPS module to obtain the geographic coordinates of the device's installation location and employs a nearest neighbor interpolation algorithm to match the climate characteristic template for the current region. The matching process utilizes data on sandstorm frequency, annual average humidity, and extreme temperatures to generate an environmental pattern weight coefficient matrix, which drives parameter initialization for the oil pressure compensation calculation module.

[0043] The temperature detection unit collects axially distributed temperature gradient data from the gearbox and uses a sliding time window to extract the temperature change rate eigenvector. This eigenvector is then fed into the piecewise fitting module of the material thermal expansion coefficient curve. The thermal expansion coefficient curve is stored in the environmental characteristic database by material grade. Upon recall, the corresponding alloy steel or cast iron material parameter table is matched based on the gearbox nameplate information. A temperature-expansion relationship surface is constructed using a cubic spline interpolation algorithm. The compensation calculation module performs a tensor product operation on the temperature gradient data and the expansion surface to generate an axial compensation coefficient matrix for the oil pressure threshold. This compensation coefficient matrix is ​​normalized and then fed into the oil pressure control equation.

[0044] The compensated threshold parameters are packaged and transmitted to the temperature-controlled oil circuit of the hydraulic locking device after CRC verification. The oil circuit control system parses the target viscosity value in the parameter package and infers the target oil temperature setpoint based on the viscosity-temperature transfer function. The heating power regulator uses a fuzzy PID control algorithm to dynamically adjust the duty cycle of the electric heating tube based on the feedback signal from the oil film temperature sensor, so that the oil viscosity fluctuation range converges to the set range. During the oil temperature control process, the pressure response delay of the hydraulic actuator is monitored in real time. If the delay exceeds the limit, a closed-loop correction instruction for the threshold parameter is triggered. The correction value is written to the historical compensation record table in the environmental characteristics database for subsequent use by the compensation calculation module.

[0045] Specifically, in the turning system of the present invention, the closed-loop feedback module performs the following steps: inputting the interlock protection false triggering event data into the sand and dust noise feature library of the environmental feature database, and updating the sliding mean filter window length; Optimizing the slope parameters of the crystal oscillator frequency compensation module based on the response delay data of the hydraulic locking device; The historical rainfall data in the environmental characteristics database is correlated to shorten the triggering interval of the hydrophobic coating self-cleaning cycle.

[0046] The closed-loop feedback module described in the present invention achieves continuous improvement in system performance through multi-source data cross-validation and parameter iterative optimization. After the interlocking protection false trigger event data is parsed by the event classifier, the time-frequency characteristics of the dust noise are extracted. The feature vector is input into the dust noise feature library of the environmental feature database for pattern matching. The successfully matched noise template drives the sliding mean filter window length adjustment strategy. The window length is dynamically shrunk in inverse proportion to the noise energy spectrum density value. The updated filter parameters are injected into the real-time data processing pipeline through the hot deployment mechanism. The location and timestamp information of the false trigger event are written into the historical operating condition log, and a noise feature distribution heat map is generated for reference in subsequent threshold optimization.

[0047] The response delay data of the hydraulic locking device is transmitted via a bus to the error analysis unit of the clock synchronization compensation module. After wavelet noise reduction processing, the delay data is input into the time series prediction model, which outputs the slope correction value of the crystal oscillator frequency compensation curve. Slope parameter optimization uses a gradient descent algorithm to iteratively adjust the compensation coefficient matrix. The adjusted parameter set is written to the control register of the FPGA programmable logic device, simultaneously triggering the closed-loop calibration process of the crystal oscillator driver circuit. During the calibration process, the time domain synchronization error of the vibration sensor is monitored simultaneously. If the error exceeds the limit, the system will fall back to the previous stable parameter version.

[0048] Historical rainfall data is extracted from the weather record sub-database of the environmental characteristics database by quarter. A sliding time window algorithm is used to calculate the rainfall frequency and intensity distribution over the past 30 days. The basic trigger interval for the self-cleaning cycle is derived based on the coating hydrophobicity attenuation model. When the number of consecutive days of rainfall exceeds a preset threshold, an exponential decay function is activated to dynamically shorten the trigger interval, with the shortening magnitude positively correlated with rainfall intensity. The adjusted cleaning cycle parameters are transmitted to the control unit of the pinion meshing mechanism via the Modbus protocol, driving the high-pressure water mist cleaning device of the hydrophobic coating to operate according to the optimized timing. A signal indicating the completion of the cleaning operation is fed back to the environmental characteristics database, updating the coating status assessment parameter table.

[0049] Specifically, in the turning system of the present invention, the dynamic weight allocation unit and the sliding mean filter unit work in coordination: When the particle concentration detection value increases by 100μg / m³, the sliding mean filter window length is shortened by 20%, and the dynamic weight allocation unit increases the weight coefficient of the vibration sensor by 0.3; The output signal of the sliding mean filter unit triggers the operating mode switching instruction of the environmental feature database.

[0050] The dynamic weight allocation unit and the sliding mean filter unit described in the present invention achieve dual anti-interference optimization through an environmental parameter coupling mechanism. The particle concentration detection unit uses a laser scattering sensor to monitor the suspended particulate matter concentration in the intake duct in real time. When the detection value increment exceeds 100 μg / m³, the sliding mean filter window length adjustment mechanism is triggered. The window length is dynamically reduced according to the concentration increment gradient using a piecewise linear function. Every 10 μg / m³ increment corresponds to a 3% reduction in the window length lower limit, to a minimum of 60% of the original length. The window length parameter is transmitted to the sliding mean filter configuration register via a ring buffer, enabling online hot update of the filter parameters.

[0051] The dynamic weight allocation unit establishes a confidence mapping model for vibration sensor and optical measurement data. When the particle concentration increment condition is met, it calls the historical data quality assessment report of the vibration sensor and increases the weight coefficient by 0.3 based on the improvement in signal-to-noise ratio. A double-buffered switching strategy is used during the weight adjustment process to smoothly transition the coefficients between data collection intervals, preventing jumps in measurement data. The adjusted weight parameters are input into the multi-source data fusion engine, and the fusion engine's output signal is synchronized with the sliding mean filter data through the timestamp alignment module.

[0052] The data stream processed by the sliding mean filter contains time-domain smoothing signatures, which trigger the operating mode switching command generation module in the environmental feature database. This command generation module analyzes the spectral entropy distribution characteristics of the data stream and dynamically time-warps the data against the dust spectrum template stored in the database. When the matching degree exceeds a threshold, it generates an operating mode switching control word. This control word is then sent via a high-speed bus to the priority scheduler in the interlocking protection decision module, driving resource reallocation within the protection threshold calculation thread.

[0053] Particulate matter concentration change data is synchronously written to the dust event record table in the environmental characteristics database. This record table associates timestamps, concentration gradients, and weight adjustment parameters to form a closed-loop optimization dataset. This optimized dataset is used to regularly update the vibration sensor's weight mapping model via an offline training platform. The updated model parameters are then downloaded over the air into the flash memory of the field device, completing the iterative upgrade of the system's anti-interference capabilities.

[0054] Specifically, in the turning system of the present invention, the climate zone index table is linked to the particulate matter concentration detection unit: When the particle concentration detection value exceeds the dust threshold preset in the environmental characteristic database, the gear meshing characteristic frequency template under the desert climate mode is called; According to the typical environmental pattern matched by the longitude and latitude coordinates, the temperature-oil pressure compensation coefficient matrix of the corresponding area is loaded.

[0055] The climate zoning index table and the particle concentration detection unit described in the present invention achieve precise matching of compensation parameters through cross-validation of geographical environmental characteristics and real-time operating conditions. The climate zoning index table is constructed based on the Köppen climate classification method and is stored in the distributed nodes of the environmental characteristics database. The index table divides climate zones according to longitude and latitude grids and associates historical sandstorm frequency, average annual humidity, and extreme temperature data. When the output value of the particle concentration detection unit exceeds the sandstorm threshold dynamically adjusted by the environmental characteristics database, the desert climate pattern recognition engine is triggered. The engine calls the desert operating condition spectrum feature set stored in the gear meshing characteristic frequency template library. The feature set includes the gear pair side clearance increase pattern caused by wind and sand wear and the high-frequency vibration noise distribution template.

[0056] The particle concentration detection unit uses a laser scattering sensor to collect real-time suspended particulate matter concentration in the air intake duct. The detection signal is transmitted via a 4-20mA current loop to the dust event analysis module in the environmental characteristics database. The dust threshold is dynamically adjusted based on historical equipment operating data. A sliding time window algorithm is used to calculate the mean and standard deviation of particle concentration over the past 30 days to generate an adaptive threshold update curve. Threshold violations trigger an operating mode switch command, which is then sent via a high-speed bus to the priority scheduler in the interlocking protection decision module.

[0057] The latitude and longitude coordinate matching module integrates a GPS positioning unit, receiving satellite signals to parse the geographic coordinates of the device's installation location. It then uses a bilinear interpolation algorithm to locate the target grid area within the climate zone index table. During the matching process, the temperature-oil pressure compensation coefficient matrix for the corresponding area is loaded. This matrix is ​​constructed dimensionally based on the material's thermal expansion coefficient and the oil's viscosity-temperature characteristics and stored in the compensation parameter sub-library within the environmental characteristics database. Using a tensor decomposition algorithm, the compensation coefficient matrix is ​​split into an axial temperature gradient compensation factor and a radial pressure distribution weight vector. These are then input into the oil pressure control equation to generate dynamic threshold adjustments.

[0058] The gear meshing characteristic frequency template for the desert climate model contains a set of characteristic frequency bands for tooth pitting caused by wind and sand erosion. During the template matching process, a dynamic time warping algorithm is used to align the real-time vibration spectrum with the template library data. When the matching degree exceeds a set threshold, the tooth backlash compensation logic is activated. The compensation logic drives the displacement sensor of the hydraulic locking device to collect axial movement. This information is combined with the temperature-oil pressure compensation coefficient matrix to generate an oil film thickness adjustment command. This command is used to adjust the lubrication pump output pressure via a PID controller. The compensated oil film pressure data is then written back to the operating condition log in the environmental characteristic database.

[0059] Specifically, in the cranking system of the present invention, the optimization of the slope of the crystal oscillator compensation curve includes: Establishing a mapping relationship model between the hydraulic response delay time and the clock synchronization accuracy; Iteratively adjusting the correction coefficient of the crystal oscillator frequency compensation module based on a gradient descent algorithm; The optimized slope parameters are written into the register group of the FPGA programmable logic device.

[0060] The slope optimization of the crystal oscillator compensation curve described in the present invention achieves a closed-loop improvement in clock synchronization accuracy through cross-system parameter coupling and hardware-level parameter updates. The hydraulic response delay time data acquisition module obtains the step response curve from the pressure sensor of the hydraulic locking device, extracts the rising edge delay time and steady-state error data, and synchronously acquires the timestamp synchronization error value output by the clock synchronization compensation module. A phase difference correlation model between the delay time and the synchronization error is constructed through time series analysis methods. The model training process uses a sliding time window to intercept data segments under continuous working conditions, calculates the integrated area of ​​the phase difference per unit time as a quantitative indicator of the mapping relationship, and generates a clock synchronization accuracy correction coefficient matrix.

[0061] The gradient descent algorithm constructs an iterative optimization path based on a loss function defined as the weighted sum of the historical RMS synchronization error and the current correction coefficient. During the algorithm initialization phase, the default correction coefficient set for the crystal oscillator frequency compensation module is loaded. During each iteration, the learning rate parameter is adjusted based on the real-time rate of change of the hydraulic response delay data. The loss function minimum is searched within the feasible solution space of the compensation coefficient matrix. The updated correction coefficients are cross-validated to prevent overfitting. The validation dataset contains crystal oscillator drift test records under different temperature gradients.

[0062] The optimized slope parameters are written to the control registers of the FPGA programmable logic device (FPGA) via the JTAG interface. The registers are configured in double-buffered mode to support online parameter hot-switching. The parameter writing process triggers a closed-loop verification process for the crystal oscillator driver circuit. The verification module compares the clock synchronization error statistics before and after compensation. Once the error converges within a threshold, the new parameter set is activated. Parameter versions that fail verification trigger a rollback mechanism, restoring the registers to the last stable state snapshot. An exception log is also generated and written to the calibration record table in the environmental feature database.

[0063] The above optimization process forms a cross-level closed-loop control chain: the response delay data of the hydraulic actuator drives the iterative training of the crystal oscillator compensation model. The optimized compensation parameters are effective in real time through hardware registers, and the calibration results are fed back to the model training module to form a self-reinforcement learning loop. The error statistics of the crystal oscillator frequency compensation module are synchronously input into the adaptive parameter adjustment module of the sliding mean filter unit, driving the coordinated optimization of the multi-level anti-interference strategy. When implemented in the turning gear of a large rotating machinery, the present invention employs a multi-source sensor network deployed at the gearbox's axial measurement points, the hub support surface, and key nodes in the transmission chain. The temperature detection unit of the environmental self-sensing sensor node uses a PT100 platinum thermal resistor, installed at 3 / 4 the axial chord length of the gear meshing surface. It monitors the tooth surface friction temperature gradient in real time, and the detection signal is transmitted to the clock synchronization compensation module via a twisted-pair shielded cable. The humidity detection unit is integrated into the annular groove surrounding the helium-neon laser emitter of the laser ranging module. It uses a capacitive sensor with a 200Hz sampling frequency to capture changes in condensation on the optical component surface. The detection data is transmitted via an RS-485 bus to the data buffer of the dynamic weight allocation unit. The particle concentration detection unit uses a MEMS laser scattering sensor, located in the front air intake duct of the laser ranging module. The detection signal is transmitted to the air curtain control unit via a 4-20mA current loop, triggering the centrifugal fan speed to adjust according to a negative exponential curve within the range of 800-2500rpm, thus establishing a dynamic equilibrium relationship between the annular air curtain flow rate and the particle concentration.

[0064] The clock synchronization compensation module runs on an embedded real-time operating system. Temperature and humidity data undergo 24-bit Σ-Δ analog-to-digital conversion and are then fed into an LSTM neural network model to predict the crystal oscillator frequency deviation and generate clock offset correction parameters for the IEEE 1588 protocol. The vibration sensor sampling timestamp calibration uses a cubic spline interpolation algorithm. The calibration reference point is dynamically selected within the -30°C to 80°C temperature range based on the measured temperature. The calibrated data stream is transmitted via a double-shielded cable with a 0.08Ω low-impedance grounding design for the outer shield to effectively suppress signal distortion caused by electromagnetic interference from the drive train. The data bus utilizes a modified CAN 2.0B protocol. Each data frame includes a 16-bit CRC checksum and an 8μs precision timestamp. Abnormal data packets trigger an automatic retransmission mechanism.

[0065] The environmental robustness data fusion module uses a climate zone index table stored in local flash memory. When the particulate matter concentration exceeds 150 μg / m³, it activates a threshold noise reduction strategy using the DB8 wavelet basis function to remove high-frequency noise components above 20 kHz. When humidity exceeds 90%, the dynamic weight allocation unit increases the magnetic encoder weight coefficient to 0.75 and applies a hysteresis compensation algorithm to compensate for the 5ms delay error in laser ranging caused by water vapor refraction. The oil moisture compensation unit combines the dielectric constant measurement value with the thermal expansion parameters of the B25 magnetic alloy to generate a nonlinear correction coefficient matrix. The compensated oil film thickness data is transmitted to the interlocking protection decision module via Modbus-TCP.

[0066] The third-order spectral kurtosis analysis unit of the interlocking protection decision module uses complex Morlet wavelets to decompose the vibration signal into 32 subbands, extracting the gear mesh characteristic frequencies and then dynamically time-warping them against the desert operating condition spectrum template. When particulate matter concentrations consistently exceed the standard, the sliding mean filter window length is adjusted by a gradient of 15% for every 50 μg / m³ concentration increment to suppress transient pulse interference. The dynamic threshold adjustment unit uses the thermal expansion coefficient curve of SAE4140 alloy and combines it with axial temperature gradient data to calculate the oil pressure compensation. The generated control command is sent to the PID controller of the hydraulic locking device via PROFIBUS-DP, achieving an oil temperature control accuracy of ±1.5°C. The closed-loop feedback module inputs hydraulic response delay data into the crystal oscillator compensation model, optimizes the slope parameters of the FPGA register using a gradient descent algorithm, and simultaneously updates the sand and dust noise feature template in the environmental feature database, forming an 8-hour parameter self-optimization cycle.

[0067] The technical features of the present invention are explained as follows: Multi-source sensor network: A heterogeneous data acquisition system consisting of multiple sensors (such as temperature, humidity, particulate matter concentration, vibration, and displacement sensors) deployed at monitoring points on the gearbox, hub, and drive chain to obtain multimodal data on the rotor's operating status and environmental parameters in real time.

[0068] Environmental self-sensing sensor node: A sensor group with integrated temperature, humidity, and particulate matter concentration detection functions, including: Temperature detection unit: A thermistor installed at the axial measuring point on the gear meshing surface to monitor the friction heat field distribution on the tooth surface.

[0069] Humidity detection unit: A capacitive sensor integrated on the periphery of the optical channel of the laser ranging module to monitor the condensation status on the surface of the optical component.

[0070] Particle concentration detection unit: A sensor based on the laser scattering principle is placed in the air intake duct to detect the concentration of suspended particulate matter in real time.

[0071] Clock Synchronization Compensation Module: This embedded processing unit, based on an environmentally coupled clock synchronization protocol, dynamically corrects sensor node clock offsets using a crystal oscillator frequency compensation algorithm. It uses temperature and humidity data to predict crystal oscillator drift and reconstructs the vibration sensor's timestamp using an interpolation calibration algorithm.

[0072] Environmentally coupled clock synchronization protocol: An improved clock synchronization mechanism that combines environmental parameters (temperature and humidity) with the IEEE 1588 protocol. It dynamically adjusts timestamp alignment accuracy by analyzing the frequency deviation characteristics of the crystal oscillator affected by temperature gradients.

[0073] Time-space synchronized data stream: A collection of multi-sensor data that has been clock-synchronized and compensated, including a unified time reference and spatial location tags, is transmitted to the data bus via a double-shielded cable to suppress electromagnetic interference.

[0074] The environmental robustness data fusion module includes the following functional units: Wavelet threshold denoising unit: Dynamically selects wavelet basis functions based on dust concentration to remove high-frequency noise components.

[0075] Dynamic weight allocation unit: adjusts the weight ratio of the magnetic encoder and laser ranging data according to the humidity detection value to compensate for the optical measurement error.

[0076] Oil moisture compensation unit: Combines the dielectric constant detection value with the database parameters to calculate the nonlinear correction coefficient of the oil film thickness.

[0077] Interlocking protection decision module: It consists of a third-order spectrum kurtosis analysis unit, a sliding mean filter unit, and a threshold dynamic adjustment unit: Third-order spectral kurtosis analysis: Extract the gear meshing characteristic frequency through complex Morlet wavelet decomposition to eliminate environmental noise interference.

[0078] Sliding mean filter: Dynamically adjust the filter window length according to the dust concentration to suppress transient pulse interference.

[0079] Dynamic adjustment of threshold: Call the material thermal expansion coefficient curve and combine it with temperature gradient data to generate the oil pressure protection threshold compensation.

[0080] Closed-loop feedback module: By collecting hydraulic actuator response delay data and interlocking event records, it reversely optimizes crystal oscillator compensation parameters and the dust noise feature library to form an adaptive parameter update mechanism. Specifically, it includes: Crystal oscillator compensation curve slope optimization: Iteratively adjust FPGA register parameters based on the gradient descent algorithm.

[0081] Dust noise feature library: stores noise spectrum templates in interlocking false triggering events, which are used to update the sliding mean filter parameters.

[0082] Climate zone index table: A database built based on the Köppen climate classification method that associates longitude and latitude coordinates with typical environmental patterns (such as desert climate) and is used to load region-specific temperature-oil pressure compensation coefficient matrices.

[0083] Temperature-controlled oil circulation circuit: A closed-loop oil temperature control system in the hydraulic locking device adjusts the heating power through the PID algorithm to maintain stable oil viscosity and dynamically adjusts the compensation parameters generated by the response threshold.

[0084] Technical logic association description: This invention collects environmental and operating status data through a multi-source sensor network. After the clock synchronization compensation module unifies the time base, the environmental robustness data fusion module performs noise reduction, weight distribution, and compensation processing to generate a fused data packet that is input into the interlocking protection decision module. This module generates control instructions based on spectral feature analysis and dynamic threshold adjustment. At the same time, the closed-loop feedback module reversely optimizes system parameters based on the execution results, forming an environmentally adaptive closed-loop control link. All technical features are linked to the database via a data bus to achieve spatiotemporal alignment and collaborative optimization of multi-source heterogeneous data, solving the problem of interlocking false triggering caused by data inaccuracy in traditional turning systems.

[0085] Crystal oscillator frequency compensation algorithm (LSTM neural network model): Based on a long short-term memory (LSTM) neural network, this algorithm predicts the impact of temperature and humidity on crystal oscillator frequency drift. By analyzing the correlation between historical temperature and humidity data and crystal oscillator frequency deviation, it dynamically generates clock offset correction values ​​compliant with the IEEE 1588 protocol, addressing the inaccuracy of traditional linear compensation models in nonlinear environments. The input layer receives time-series temperature and humidity data, the hidden layer extracts the nonlinear characteristics of environmental parameters and crystal oscillator drift, and the output layer generates clock offset compensation coefficients. Model training uses a sliding time window to capture continuous operating data, and the loss function is defined as the mean squared error between the predicted and actual frequency deviations.

[0086] Wavelet Threshold Denoising Model (db8 wavelet basis function): This model performs multi-scale decomposition of the torque sensor signal and dynamically adjusts the high-frequency subband coefficients based on dust concentration using a soft thresholding strategy to remove transient noise caused by particle collisions. The Daubechies 8 wavelet basis function is used to decompose the signal into five layers, dynamically adjusting the threshold range for the high-frequency coefficients in each layer based on particle concentration. When dust concentration exceeds the limit, high-frequency noise suppression is enhanced to preserve signal energy in the characteristic frequency band of gear meshing.

[0087] A third-order spectral kurtosis analysis model (complex Morlet wavelet transform) extracts gear meshing characteristic frequencies from the vibration signal. The spectral kurtosis of each frequency band is calculated to identify the characteristic frequency band of the fault and suppress ambient vibration and noise interference. Complex Morlet wavelet basis functions are used to decompose the vibration signal into 32 frequency bands. The third-order spectral kurtosis of each subband is calculated, and a kurtosis-frequency matrix is ​​constructed. A peak search algorithm is used to select frequency bands with kurtosis values ​​exceeding a threshold to generate a gear health assessment index.

[0088] Dynamic Weight Allocation Model (Humidity-Related Optical Error Compensation): When humidity exceeds 90%, the error correlation model between the magnetic encoder and the laser ranging module dynamically adjusts data fusion weights to compensate for laser ranging errors caused by water vapor refraction. A humidity-optical refractive index mapping table is established, and the accumulated error compensation for laser ranging is calculated based on real-time humidity values. The weight allocation coefficients are updated using a sliding window mechanism, with the window length adaptively adjusted based on the rate of humidity change. In heavy rain conditions, the window is shortened to 5 seconds to quickly respond to sudden changes.

[0089] Oil Moisture Compensation Model (Non-Linear Correction Coefficient Calculation): This model combines the dielectric constant measured by the oil film thickness sensor with the oil moisture compensation parameters in the database to generate a non-linear correction coefficient to compensate for the effect of water content on the dielectric properties. The oil moisture content-dielectric constant curve from the environmental characteristics database is used to fit the correction coefficient to the current operating conditions using the least squares method. A temperature gradient compensation factor is introduced, and the correction coefficient is recalibrated based on the gearbox axial temperature distribution.

[0090] Sliding mean filter model (window adjustment associated with dust concentration): Dynamically adjusts the filter window length based on the rate of change of particle concentration, suppressing transient pulse interference caused by dust impact while preserving the effective characteristics of the gear torque signal. An exponential decay function relationship is established between concentration increment and window length, with the lower limit of the window length shortened by 15% for every 50 μg / m³ concentration increment. The filtered output signal passes through a phase compensator to eliminate delay errors and ensure timing alignment with the threshold decision module.

[0091] Dynamic Threshold Adjustment Model (Coupling Thermal Expansion Coefficient Calculation): This model uses the material thermal expansion coefficient curve and axial temperature gradient data to calculate the oil pressure protection threshold compensation, addressing threshold drift caused by metal deformation under low-temperature operating conditions. Based on the thermal expansion coefficient curve of SAE4140 alloy, a piecewise linear interpolation algorithm is used to construct a temperature-expansion relationship surface. The compensation coefficient matrix is ​​generated through a tensor product operation, normalized, and then input into the oil pressure control equation.

[0092] Closed-loop feedback optimization model (gradient descent algorithm): Based on the mapping relationship between hydraulic response delay data and clock synchronization error, the slope parameters of the crystal oscillator frequency compensation module are iteratively optimized to improve clock synchronization accuracy. A loss function is defined as the weighted sum of the historical RMS value of synchronization error and the current correction coefficient. A gradient descent algorithm is used to search for optimal parameters within the feasible solution space. The optimized parameters are written to the FPGA register bank via the JTAG interface, supporting online hot switching and exception rollback.

[0093] Climate Zoning Matching Model (Köppen Climate Classification): This model matches the climate zoning index table based on longitude and latitude coordinates, loading a region-specific temperature-oil pressure compensation coefficient matrix to achieve adaptive threshold adjustment for the geographic environment. A bilinear interpolation algorithm is used to locate the target grid area in the index table. Parameters such as sandstorm frequency and average annual humidity are used to generate an environmental pattern weight matrix, which drives the initialization of the oil pressure compensation calculation module.

[0094] Data acquisition and synchronization: A multi-source sensor network collects environmental and mechanical signals, and the clock synchronization compensation module uses the LSTM model to unify the time base and generate a spatiotemporally synchronized data stream.

[0095] Environmental interference suppression: The wavelet threshold denoising model works in conjunction with the dynamic weight allocation model to remove noise from the frequency domain and weight allocation levels, respectively, to enhance data credibility.

[0096] Fault feature extraction and protection decision-making: The third-order spectral kurtosis analysis model extracts the gear meshing characteristic frequency, the sliding mean filter model suppresses transient interference, and the threshold dynamic adjustment model generates environmentally adaptive protection thresholds.

[0097] Closed-loop optimization and parameter update: The closed-loop feedback model optimizes the crystal oscillator compensation parameters through the gradient descent algorithm, and the climate zoning matching model loads the geographical environment characteristic parameters to form a global adaptive control link.

[0098] By leveraging the synergistic effects of these models, this invention addresses the interlocking mistriggering issue in conventional turning systems caused by temporal and spatial misalignment of multi-source data. Each model dynamically adjusts its processing strategy based on environmental parameters, enabling full-process adaptive control from data acquisition, interference suppression, feature extraction, to closed-loop optimization. This significantly improves interlocking protection accuracy and system reliability.

[0099] This invention addresses the problem of sensor data time base misalignment through a spatiotemporal synchronization compensation mechanism within a multi-source sensor network. Environmental self-sensing sensor nodes collect temperature, humidity, and particulate matter concentration data, and a clock synchronization compensation module dynamically corrects sensor clock offsets based on an environmental coupling protocol. Specifically, temperature and humidity signals are fed into a crystal oscillator frequency compensation algorithm to construct a time-base error model for low-temperature crystal oscillator drift and high-temperature thermal noise. Vibration sensor timestamps are reconstructed through an interpolation calibration algorithm to generate a spatiotemporal synchronized data stream. Calibration data is transmitted via a double-shielded cable, suppressing electromagnetic interference and ensuring that multi-source data is transmitted to the data bus under a unified time base, eliminating millisecond-level timestamp deviations.

[0100] At the data fusion level, the environmental robustness data fusion module utilizes compensation parameters from the environmental feature database to optimize data credibility under environmental interference. The wavelet threshold noise reduction unit dynamically removes high-frequency noise based on particulate matter concentration. The dynamic weight allocation unit increases the weight of the magnetic encoder and decreases the weight of the optical measurement when humidity exceeds the standard to compensate for water vapor refraction errors. The oil moisture compensation unit combines dielectric constant measurements with database parameters to calculate nonlinear correction coefficients. Multi-source data is fused using a weighted Kalman filter to generate a decision data packet. Spatial and temporal alignment is used to eliminate correlation failures caused by differences in the spatial distribution of sensor data.

[0101] The interlocking protection decision module and closed-loop feedback mechanism collaborate to suppress false triggering. Third-order spectral kurtosis analysis extracts the characteristic frequencies of gear meshing, and the sliding mean filter dynamically adjusts the window length based on dust concentration to suppress transient pulse interference. The threshold dynamic adjustment unit utilizes material thermal expansion coefficient and temperature gradient data to calculate the oil pressure protection threshold compensation and generate hydraulic control instructions. The closed-loop feedback module feeds execution delay data back to the clock synchronization module to optimize crystal oscillator compensation parameters. Interlocking false triggering events drive updates to the dust noise signature library, and historical rainfall data shortens the self-cleaning cycle trigger interval. This creates an environmental parameter-driven threshold adaptive convergence mechanism, reducing the risk of protection logic misjudgments.

Claims

1. A turning system, characterized in that: include: A multi-source sensor network, deployed at the gearbox, hub, and drive train monitoring points, includes environmental self-sensing sensor nodes and vibration and displacement sensors. The environmental self-sensing sensor nodes output temperature, humidity, and particulate matter concentration detection signals to collect multimodal data on the dynamic operation of the rotor; A clock synchronization compensation module receives temperature and humidity detection signals from the multi-source sensor network, dynamically corrects the clock offset of the sensor nodes through an environmental coupling clock synchronization protocol, generates a time-space synchronization data stream, and outputs it to a data bus; an environmental robustness data fusion module, which receives a spatiotemporally synchronized data stream from the data bus, calls compensation parameters from an environmental feature database stored in a local memory, performs adaptive filtering and noise reduction on the torque sensor signal, performs dielectric constant compensation on the oil film thickness sensor data, and dynamically allocates weight coefficients of multi-source data according to real-time environmental conditions, generating a fused data packet and transmitting it to the decision bus; An interlocking protection decision module obtains a fused data packet from the decision bus, extracts fault characteristic frequencies based on third-order spectral kurtosis analysis, matches the environmental operating mode, dynamically adjusts the gearbox oil pressure protection threshold, and generates control instructions for the hydraulic locking device and the pinion meshing mechanism; The closed-loop feedback module collects the response delay data of the hydraulic locking device and the interlock protection trigger record, feeds the response delay data back to the crystal oscillator frequency compensation algorithm of the clock synchronization compensation module, writes the interlock protection trigger record into the historical operating condition log of the environmental feature database, and optimizes the drive system parameters online.

2. The barring system according to claim 1, characterized in that: The environment self-sensing sensor node includes: The temperature detection unit is deployed at the axial measurement point of the gear meshing surface and outputs the temperature detection signal to the clock synchronization compensation module; The humidity detection unit is integrated in the periphery of the optical channel of the laser ranging module and outputs the humidity detection signal to the dynamic weight distribution unit; The particle concentration detection unit is connected to the centrifugal fan control terminal of the air curtain protection layer deployed outside the laser ranging module; The output signal of the particle concentration detection unit triggers the centrifugal fan speed adjustment through the PID regulator of the air curtain control unit, thereby maintaining the air flow velocity of the air curtain protective layer in a negative correlation with the particle concentration.

3. The barring system according to claim 1, characterized in that: The clock synchronization compensation module performs: Input the signals from the temperature detection unit and the humidity detection unit into the crystal oscillator frequency compensation algorithm to calculate the clock offset correction value of the IEEE 1588 protocol; Based on the time base error model of low-temperature crystal oscillator drift and high-temperature thermal noise, the sampling timestamps of vibration sensors in multi-source sensor networks are interpolated and calibrated. The calibrated data stream is transmitted to the data bus via a double-shielded cable, and the grounding impedance of the outer shielding layer of the cable is ≤0.1Ω.

4. The barring system according to claim 1, characterized in that: The environmental robustness data fusion module includes: a wavelet threshold noise reduction unit, activating a high-frequency noise stripping operation in response to an output signal of the particle concentration detection unit; Dynamic weight allocation unit: when the detection value of the humidity detection unit is greater than 90%, the data weight coefficient of the magnetic encoder is increased to 0.75, and the data weight coefficient of the laser ranging module is reduced to 0.25; The oil moisture compensation unit receives the dielectric constant detection value of the oil film thickness sensor and calculates the nonlinear correction coefficient in combination with the oil moisture compensation parameter in the environmental characteristic database.

5. The barring system according to claim 1, characterized in that: The interlocking protection decision module includes: A third-order spectral kurtosis analysis unit to extract the gear meshing characteristic frequency from the vibration sensor signal of the multi-source sensor network; The sliding mean filter unit activates the sliding window length adaptive adjustment strategy when the particle concentration detection value exceeds 200μg / m³ for 30 seconds; The threshold dynamic adjustment unit calls the material thermal expansion coefficient curve stored in the environmental feature database, combines the gradient data of the temperature detection unit, and dynamically calculates the oil pressure protection threshold compensation amount.

6. The barring system according to claim 5, characterized in that: The threshold dynamic adjustment unit performs: Calling the climate zone index table in the environmental characteristics database to match typical environmental patterns according to latitude and longitude coordinates; Calculate the oil pressure threshold compensation based on the gradient data of the medium temperature detection unit and the thermal expansion coefficient curve of the gearbox material; The compensated threshold parameter is input into the temperature control circulation oil circuit of the hydraulic locking device.

7. The barring system according to claim 1, characterized in that: The closed-loop feedback module executes: inputting the interlock protection false trigger event data into the sand and dust noise feature library of the environmental feature database, and updating the sliding mean filter window length; Optimizing the slope parameters of the crystal oscillator frequency compensation module based on the response delay data of the hydraulic locking device; The historical rainfall data in the environmental characteristics database is correlated to shorten the triggering interval of the hydrophobic coating self-cleaning cycle.

8. The barring gear system according to claim 4 or 5, characterized in that: The dynamic weight allocation unit works in conjunction with the sliding mean filtering unit: When the particle concentration detection value increases by 100μg / m³, the sliding mean filter window length is shortened by 20%, and the dynamic weight allocation unit increases the weight coefficient of the vibration sensor by 0.3; The output signal of the sliding mean filter unit triggers the operating mode switching instruction of the environmental feature database.

9. The barring gear system according to claim 2 or 5, characterized in that: The climate zone index table is linked to the particle concentration detection unit: When the particle concentration detection value exceeds the dust threshold preset in the environmental characteristic database, the gear meshing characteristic frequency template under the desert climate mode is called; According to the typical environmental pattern matched by the longitude and latitude coordinates, the temperature-oil pressure compensation coefficient matrix of the corresponding area is loaded.

10. The barring gear system according to claim 3 or 7, characterized in that: The crystal oscillator compensation curve slope optimization includes: Establishing a mapping relationship model between the hydraulic response delay time and the clock synchronization accuracy; Iteratively adjusting the correction coefficient of the crystal oscillator frequency compensation module based on a gradient descent algorithm; The optimized slope parameters are written into the register group of the FPGA programmable logic device.

Images