An experimental calibration method based on multi-sensor signal fusion processing

Abstract

The present invention discloses an experimental calibration method based on multi-sensor signal fusion processing, which specifically relates to the field of data analysis, including a step-by-step anti-disturbance protocol, data stream processing and real-time analysis, and multi-level logical judgment and calibration execution. The present invention pioneered a step-by-step anti-disturbance protocol, which stimulates and corrects the potential errors of sensors in real time without interrupting system operation, completely replacing the traditional shutdown calibration mode; constructs a multi-level disturbance defense system to effectively suppress complex interference such as spectrum aliasing, overload distortion, and signal-to-noise ratio attenuation caused by strong coupling of electromechanical and thermal multi-physical fields; integrates laser spatial calibration and dynamic stiffness modulation technology to accurately diagnose and compensate for spatial drift errors caused by deformation of the mounting base and mechanical loosening. Ultimately, the long-term accuracy stability and data fusion reliability of multi-source sensors under complex working conditions are achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of data analysis, and more particularly to an experimental calibration method based on multi-sensor signal fusion processing. Background Art

[0002] In the condition monitoring and fault diagnosis of large rotating machinery (such as turbines, compressors, and generator sets), existing technologies use periodic shutdown calibration combined with a single-dimensional compensation strategy for sensor calibration. The standard process is: manually calibrate the sensor using standard measuring instruments when the system is in a static state, or apply a preset compensation coefficient (such as the temperature drift compensation curve of a thermistor) to correct the sensor output based on feedback from a single environmental factor sensor during operation. The calibration process relies on static data collection and manual parameter adjustment under fixed operating conditions.

[0003] However, existing methods have serious limitations: first, the shutdown calibration mode destroys the continuous operation state of the system, cannot capture transient process errors, and has high maintenance costs; second, a single environmental factor compensation mechanism is difficult to cope with multi-physical field coupling disturbances such as voltage fluctuations, mechanical shocks, electromagnetic interference, and oil contamination, resulting in the failure of compensation parameters under dynamic working conditions; third, there is a lack of detection capabilities for deep coupling errors such as sensor mounting base deformation, loose mechanical connections, and spatial position drift, resulting in the continuous accumulation of complex errors such as vibration transmission path distortion and electromagnetic induction crosstalk, which cannot be corrected. Summary of the Invention

[0004] In order to overcome the above-mentioned defects of the prior art, the present invention provides an experimental calibration method based on multi-sensor signal fusion processing, which solves the problems raised in the above-mentioned background technology through the following scheme.

[0005] To achieve the above objectives, the present invention provides the following technical solution: an experimental calibration method based on multi-sensor signal fusion processing, comprising:

[0006] S1: Step-by-step counter-disturbance protocol: Design a multi-stage electromechanical and thermal coupling disturbance scenario to stimulate potential sensor errors;

[0007] S2: Data stream processing and real-time analysis: Dynamic benchmarks are generated through optical and material labeling technology to replace traditional manual calibration. Data flows through a spatiotemporal alignment pipeline to unify sampling rates and coordinate systems. Online residual analysis calculates the deviation between multi-source signals and the physical model. Frequency domain coherence monitoring verifies the correlation strength between sensors.

[0008] S3: Multi-level logical judgment and calibration execution: Implement a four-level decision rule base and dynamic confidence weight management to diagnose sensor status and trigger calibration.

[0009] Preferably, the first stage of the multi-stage electromechanical thermal coupling disturbance scenario in S1 is the establishment of basic working conditions and low-frequency disturbance injection; the second stage is load step and high-frequency common-mode interference; the third stage is transient impact and dynamic stiffness modulation; and the fourth stage is surge boundary limit test.

[0010] Preferably, the basic working condition establishment and low-frequency disturbance injection first maintain the rated speed for ten minutes in the no-load state of the equipment, collect sensor baseline data of vibration, temperature, current and oil parameters, and then inject the first-level disturbance: superimpose ±5% 0.5 times the power frequency voltage fluctuation on the power input side through the programmable frequency converter, and the fluctuation frequency is 25Hz, which lasts for five minutes. The first stage focuses on monitoring the harmonic spectrum changes of the current transformer and the response delay of the low-frequency component of the bearing seat vibration, where the low-frequency component is the frequency band below 200Hz, and synchronously records the temperature distribution uniformity of the stator core surface with the infrared thermal imager. When it is detected that the vibration phase lag exceeds 15° or the temperature change of the stator core surface within an axial distance of more than 8°C per meter is detected, the sensor gain calibration module is triggered.

[0011] Preferably, the sensor gain calibration module first collects the output signal S of the target sensor under the reference working condition. base At the same time, the real physical quantity P at the same position is obtained by laser Doppler vibrometer ref , build a gain correction model based on deviation analysis: , G cal is the calibration gain coefficient, k is the temperature compensation factor, which is calculated in real time by the adjacent temperature sensor, μ offset Indicates the signal DC bias, S raw Represents the original output signal of the sensor, μ noise The execution process adopts a three-step closed-loop strategy to calculate the mean value of high-frequency noise. First, a sinusoidal sweep signal with increasing amplitude is injected in the frequency range of 5Hz-500Hz to establish the sensor transfer function curve. Second, the gain-frequency characteristic surface is fitted by the least squares method. Finally, the correction coefficient k is written into the sensor transmitter register.

[0012] Preferably, the load step and high-frequency common-mode interference increase the equipment load step to 80% of the rated value, and after the operation is stable, the second-level disturbance is injected: a high-frequency common-mode interference signal with a frequency of 5kHz and a peak-to-peak value of 50Vpp is injected into the control loop for 120 seconds. In the second stage, three operations are performed simultaneously: a high-voltage differential probe is used to capture the transient voltage waveform of the motor terminal to identify the electromagnetic interference intrusion path; a laser Doppler vibrometer is turned on to scan the rotor journal area to quantify the signal-to-noise ratio attenuation curve under high-frequency interference; ISO 4406 grade 16 standard particles are injected into the lubricating oil circuit, and the particle size distribution range is 15-25μm. The fidelity of the pollution working condition signal is verified by correlation analysis of the electrostatic sensor and high-speed microscopy. When the measurement error of the current harmonic sensor at the 5kHz frequency point exceeds ±3%, the anti-aliasing filter reconstruction algorithm is automatically activated.

[0013] Preferably, the anti-aliasing filter reconstruction algorithm adopts a dual-path adaptive filtering architecture: the main path retains the original signal x(n), the auxiliary path generates the reconstructed signal y(n), and the weight vector is dynamically adjusted through the LMS filter: , where M is the filter order, the default value is 128, μ is the convergence factor, the dynamic range is 0.01-0.001, e(n) is the error signal, the reference source uses the output of the fiber optic current sensor, and the reconstruction process is divided into three stages: the initialization stage establishes the out-of-band attenuation model, the stopband attenuation is greater than 80dB, and the convergence stage monitors the stopband energy ratio in real time , E stopband is the stopband energy, E passband is the passband energy, and the target value is less than 10 -4 , the optimal weight vector is locked in the stable stage, and the final output signal is resampled by 128 times interpolation to ensure that the Nyquist frequency is extended to four times the original sampling rate, completely eliminating the spectrum aliasing phenomenon.

[0014] Preferably, the transient impact and dynamic stiffness modulation inject a third-level disturbance while maintaining an 80% load condition: a current spike with a peak value of 200A and a pulse width of 100μs is generated by a preset short-circuit device, and a pneumatic exciter is synchronously triggered to apply a 3ms half-sine mechanical impact with an acceleration amplitude of 15 times the acceleration of gravity. A high-speed acquisition card with a sampling rate of 100MS / s is used to capture the overload recovery characteristics of the sensor, requiring that the time for the signal to recover to the normal fluctuation band does not exceed 50ms, where the normal fluctuation band is defined as the range of ±2 times the standard deviation. At the same time, the base is hydraulically released by 50% of the preload force, so that the natural frequency drops from 500Hz to 200Hz. The hydraulic release of the base preload takes 3 minutes, and the equipment foundation displacement is monitored in real time by a laser tracker to construct a vibration transfer function matrix: ,in represents the new transfer function, Vib bearing Indicates the bearing vibration spectrum, Vib baseRepresents the basic vibration spectrum, J is the imaginary unit, which is the primitive of the negative domain, and the physical meaning of -J is phase lag. is the phase angle. When the transfer function amplitude changes by more than 20 dB or the phase shift exceeds 30°, the sensor installation coupling is determined to be invalid and the recalibration process is started.

[0015] Preferably, the recalibration process first deploys the laser tracker to establish a reference grid with a grid spacing of 50mm in the device base coordinate system, and then performs three-dimensional space mapping, including static calibration, dynamic tracking, and coordinate transformation; the static calibration uses a robotic arm to drive the reflective target ball to traverse each sensor installation point and record the position matrix , x i ,y i , z i Represents the X, Y, and Z axis coordinate values ​​of the sensor in the basic coordinate system. The dynamic tracking captures the vibration trajectory Q(t) of the sensor housing at 2,000 frames per second through CCD high-speed camera under 80% load conditions. The coordinate transformation solves the homogeneous transformation matrix , where R is the rotation matrix and t is the translation vector. After the transformation is completed, the spatial compensation vector is generated ,Write spatial coordinate synchronization unit,δ x , δ y , δ z is the linear displacement compensation, δ θx , δ θy , δ θz is the rotation compensation angle around the X / Y / Z axis.

[0016] Preferably, the surge boundary limit test closes the outlet valve to a critical opening range of 65%-70% to create a progressive surge condition: the pressure relief valve is periodically adjusted at a frequency of 0.2 Hz, with its opening changing in steps between 30% and 80%; the flow channel pulsation wavefront propagation velocity is captured by a dynamic pressure sensor; the turning point of the axial to radial energy ratio is analyzed using a vibration accelerometer; and the temperature entropy value of the inlet guide vane area is calculated based on infrared thermal imager data: , where S T is the temperature entropy, p i The distribution probability of each temperature range is represented. The fusion system needs to trigger a surge warning when the temperature entropy sudden change growth rate exceeds 0.8 bits / s and the pressure pulsation amplitude exceeds 25% of the rated value. At the same time, it needs to verify that the multi-sensor time alignment error is less than 100 μs.

[0017] Preferably, the data stream processing of S2 is based on the raw signal acquisition of the sensor array. All data are time-stamped based on the IEEE 1588 precision time protocol. The raw signal is first temporarily stored in a circular buffer, and the buffer depth is set to twice the maximum expected delay. Then, the resampling engine performs spatiotemporal normalization processing: the sampling rate of each sensor is unified to a 100kHz benchmark through a cubic spline interpolation algorithm, while compensating for the microsecond delay caused by the difference in transmission cable length. In the coordinate system conversion stage, the laser tracker scans the key reference points of the device every 30 seconds and constructs the spatial transformation matrix in real time: , where Δx, Δy, Δz are the position offsets, and θ is the installation deflection angle. The converted data forms a spatiotemporal synchronization cube and is input into the fusion core for multi-dimensional analysis.

[0018] Preferably, the core of the real-time analysis in S2 is the residual monitoring engine, which calculates the multi-source residual matrix every 200 milliseconds: , i represents the sensor number, j corresponds to the physical model type, represents the real-time measurement value of the i-th sensor, represents the predicted value of the jth type of physical model. The vibration residual analysis uses the Campbell diagram comparison technology to match the measured acceleration spectrum with the critical speed trajectory output by the rotor dynamics model; the temperature residual is obtained by solving the three-dimensional heat conduction equation Generate theoretical temperature field, k is thermal conductivity, T is temperature, ρ is density, c p is the specific heat capacity, t is the time, is the gradient operator, When the residual kurtosis value exceeds 3.0 or five consecutive sampling points exceed the ±3σ boundary, the anomaly mark is triggered in real time and the root cause analysis thread is started.

[0019] Preferably, the S2 uses frequency domain coherence monitoring as an auxiliary verification means to track the transfer characteristics of current and vibration: , is the cross power spectrum density of current and vibration, and They are respectively the autopower spectrum. In the characteristic frequency band, when the coherence coefficient is lower than 0.7 and lasts for three seconds, it is determined that the sensor is inaccurate.

[0020] Preferably, the processing process adopts a pipeline architecture to ensure real-time performance: the first-level FPGA completes timestamp alignment and filtering, the second-level GPU cluster performs model calculation and residual analysis, and the third-level CPU implements logical decision-making.

[0021] Preferably, the multi-level logic judgment of S3 starts with the input of raw data, and first performs a single sensor self-test: by analyzing the signal baseline stability for 10 consecutive minutes, the standard deviation is less than 0.1% of the full scale, the signal-to-noise ratio is greater than 60dB, and the power supply ripple is less than 5mV to achieve preliminary screening. The sensor that passes the self-test enters the physical consistency check stage. At this time, the system calls the preset physical model library and compares the sensor measured value with the model predicted value in real time: the vibration signal must meet , represents the frequency domain amplitude of the measured vibration acceleration signal, σ represents the standard deviation, where By transfer function and electromagnetic force Calculated theoretical value, F em represents the electromagnetic excitation force, k t represents the motor torque constant, r represents the rotor equivalent radius, Represents the current harmonic component, and the temperature distribution must conform to the heat conduction equation The numerical solution boundary of k is thermal conductivity, T is temperature, ρ is density, c p is the specific heat capacity, t is the time, is the gradient operator, is the symbol of the partial derivative.

[0022] Preferably, when the physical consistency check fails, the system automatically activates the model parameter correction module and updates the transfer function based on the residual gradient descent method. Or the thermal conductivity k of the material, enter the multi-sensor mutual verification link through the inspection data, and cross-compare the complex coherence function of the current harmonics and vibration spectrum :like If the vibration residual exceeds the standard, it is determined to be sensor drift. And the vibration residual exceeds the standard, triggering the equipment protection mechanism.

[0023] Preferably, the sensor that fails the mutual verification will start the fourth level of confrontation test verification: reproduce the step disturbance under controlled working conditions. If the fault characteristics are reproduced, the sensor is confirmed to be failed, the data is automatically isolated and the replacement instruction is triggered; if it is not reproduced, it is determined to be a false alarm, and the confidence weight W of the sensor is adjusted. i (t), confidence management adopts a dynamic weighting model: , R recent is the residual compliance rate in the last five minutes, R history Based on the exponential decay function calculation of the number of historical failures, E env E is the environmental severity factor, when the temperature is >80℃ env The value is 0.8, and the sensor data with a final confidence level lower than 0.6 will be isolated.

[0024] Preferably, the calibration execution stage of S3 first identifies the drift mode: performs the Compensation, μ residual Represents the statistical mean of the residual sequence; the gain drift is expressed as , S raw Represents the original sensor signal, and the k value is calculated by back-calculating the adjacent sensors; nonlinear distortion activates the pre-trained neural network corrector. After calibration, a 0.5Hz sine sweep signal is immediately injected to verify phase consistency and start 24-hour Allan variance monitoring. To ensure long-term stability, , τ represents the duration of the data packet, represents the arithmetic mean of the sensor output values ​​in the kth time window, It represents the arithmetic mean of the adjacent next time window. Finally, the theoretical value of the bearing reaction force is compared with the measured value through the laser Doppler vibrometer to complete the closed-loop verification.

[0025] Technical effects and advantages of the present invention:

[0026] 1. First, based on a step-by-step anti-disturbance protocol and an online closed-loop correction mechanism, it can stimulate and capture various potential errors in real time during continuous operation, completely replacing the traditional shutdown calibration mode and significantly improving system availability and calibration timeliness.

[0027] 2. Secondly, through multi-level disturbance scenario design and multi-sensor mutual verification logic, a comprehensive defense system for strong coupling interference of electromechanical, thermal and multi-physical fields has been constructed for the first time. This effectively suppresses complex errors such as spectrum aliasing caused by voltage fluctuations, overload distortion caused by mechanical shock, and signal-to-noise ratio degradation caused by oil contamination.

[0028] 3. Finally, by creatively integrating spatial position calibration and coupling error decoupling technology, laser grid positioning and transfer function analysis are used to accurately diagnose spatial drift errors such as installation base deformation and loose mechanical connections. Dynamic stiffness modulation is used to achieve autonomous reconstruction of the measurement benchmark, fundamentally ensuring the spatial consistency and long-term stability of multi-source data fusion. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1 It is a schematic diagram of the overall structure of the present invention.

[0030] Figure 2 Schematic diagram of the structure of the step-by-step anti-disturbance protocol of the present invention.

[0031] Figure 3 This is a schematic diagram of the data stream processing and real-time analysis structure of the present invention.

[0032] Figure 4 Schematic diagram of the multi-level logic judgment and calibration execution structure of the present invention. DETAILED DESCRIPTION

[0033] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments 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 efforts are within the scope of protection of the present invention.

[0034] refer to Figures 1-4 An experimental calibration method based on multi-sensor signal fusion processing is shown, comprising:

[0035] S1: Stepwise counter-disturbance protocol: Design a multi-stage electromechanical-thermal coupled disturbance scenario to stimulate potential sensor errors.

[0036] The first stage of the multi-stage electromechanical thermal coupling disturbance scenario in S1 is the establishment of basic working conditions and low-frequency disturbance injection; the second stage is load step and high-frequency common-mode interference; the third stage is transient impact and dynamic stiffness modulation; and the fourth stage is surge boundary limit test.

[0037] The basic working condition establishment and low-frequency disturbance injection first maintain the rated speed for ten minutes in the no-load state of the equipment, collect sensor baseline data of vibration, temperature, current and oil parameters, and then inject the first-level disturbance: superimpose ±5% 0.5 times the power frequency voltage fluctuation on the power input side through the programmable frequency converter. The fluctuation frequency is 25Hz and lasts for five minutes. The first stage focuses on monitoring the harmonic spectrum changes of the current transformer and the response delay of the low-frequency component of the bearing seat vibration, where the low-frequency component is the frequency band below 200Hz, and synchronously records the temperature distribution uniformity of the stator core surface with an infrared thermal imager. When the vibration phase lag exceeds 15° or the temperature change of the stator core surface exceeds 8°C per meter along the axial direction is detected, the sensor gain calibration module is triggered.

[0038] The sensor gain calibration module first collects the output signal S of the target sensor under the reference working condition. base At the same time, the real physical quantity P at the same position is obtained by laser Doppler vibrometer ref , build a gain correction model based on deviation analysis: , G cal is the calibration gain coefficient, k is the temperature compensation factor, which is calculated in real time by the adjacent temperature sensor, μ offset Indicates the signal DC bias, S raw Represents the original output signal of the sensor, μ noiseThe execution process adopts a three-step closed-loop strategy: first, a sinusoidal sweep signal with increasing amplitude is injected in the frequency range of 5Hz-500Hz to establish the sensor transfer function curve; second, the gain-frequency characteristic surface is fitted by the least squares method; finally, the correction coefficient k is written into the sensor transmitter register.

[0039] The load step and high-frequency common-mode interference increase the equipment load step to 80% of the rated value. After the operation stabilizes, the second-level disturbance is injected: a high-frequency common-mode interference signal with a frequency of 5kHz and a peak-to-peak value of 50Vpp is injected into the control loop for 120 seconds. In the second stage, three operations are performed simultaneously: a high-voltage differential probe is used to capture the transient voltage waveform of the motor terminal to identify the electromagnetic interference intrusion path; a laser Doppler vibrometer is turned on to scan the rotor journal area to quantify the signal-to-noise ratio attenuation curve under high-frequency interference; ISO 4406 grade 16 standard particles with a particle size distribution range of 15-25μm are injected into the lubrication oil circuit, and the fidelity of the pollution condition signal is verified by correlation analysis between electrostatic sensors and high-speed microscopy. When the measurement error of the current harmonic sensor at the 5kHz frequency point exceeds ±3%, the anti-aliasing filter reconstruction algorithm is automatically activated.

[0040] The anti-aliasing filter reconstruction algorithm adopts a dual-path adaptive filtering architecture: the main path retains the original signal x(n), and the auxiliary path generates the reconstructed signal y(n). The weight vector is dynamically adjusted through the LMS filter: , where M is the filter order, the default value is 128, μ is the convergence factor, the dynamic range is 0.01-0.001, e(n) is the error signal, the reference source uses the output of the fiber optic current sensor, and the reconstruction process is divided into three stages: the initialization stage establishes the out-of-band attenuation model, the stopband attenuation is greater than 80dB, and the convergence stage monitors the stopband energy ratio in real time , E stopband is the stopband energy, E passband is the passband energy, and the target value is less than 10 -4 , the optimal weight vector is locked in the stable stage, and the final output signal is resampled by 128 times interpolation to ensure that the Nyquist frequency is extended to four times the original sampling rate, completely eliminating the spectrum aliasing phenomenon.

[0041] The transient shock and dynamic stiffness modulation injects a third-level disturbance while maintaining an 80% load condition: a current spike with a peak value of 200A and a pulse width of 100μs is generated through a preset short-circuit device, and a pneumatic exciter is synchronously triggered to apply a 3ms half-sine mechanical shock with an acceleration amplitude of 15 times the acceleration of gravity. A high-speed acquisition card with a sampling rate of 100MS / s is used to capture the sensor overload recovery characteristics. The time required for the signal to recover to the normal fluctuation band is no more than 50ms, where the normal fluctuation band is defined as the range of ±2 times the standard deviation. At the same time, 50% of the preload of the base is hydraulically released, causing the natural frequency to drop from 500Hz to 200Hz. The hydraulic release of the base preload takes 3 minutes. The foundation displacement of the equipment is monitored in real time by a laser tracker to construct the vibration transfer function matrix: ,in represents the new transfer function, Vib bearing Indicates the bearing vibration spectrum, Vib base Represents the basic vibration spectrum, J is the imaginary unit, which is the primitive of the negative domain, and the physical meaning of -J is phase lag. is the phase angle. When the transfer function amplitude changes by more than 20 dB or the phase shift exceeds 30°, the sensor installation coupling is determined to be invalid and the recalibration process is started.

[0042] The recalibration process first deploys the laser tracker to establish a reference grid with a grid spacing of 50mm in the device base coordinate system, and then performs three-dimensional space mapping, including static calibration, dynamic tracking, and coordinate transformation; the static calibration uses a robotic arm to drive the reflective target ball to traverse each sensor installation point and record the position matrix , x i ,y i , z i Represents the X, Y, and Z axis coordinate values ​​of the sensor in the basic coordinate system. The dynamic tracking captures the vibration trajectory Q(t) of the sensor housing at 2,000 frames per second through CCD high-speed camera under 80% load conditions. The coordinate transformation solves the homogeneous transformation matrix , where R is the rotation matrix and t is the translation vector. After the transformation is completed, the spatial compensation vector is generated , write the spatial coordinate synchronization unit, δ x , δ y , δ z is the linear displacement compensation, δ θx , δ θy , δ θz is the rotation compensation angle around the X / Y / Z axis.

[0043] The surge boundary limit test closes the outlet valve to a critical opening range of 65%-70% to create a progressive surge condition: the pressure relief valve is periodically adjusted at a frequency of 0.2 Hz, with its opening changing in steps between 30% and 80%. The propagation velocity of the flow channel pulsation wavefront is captured using a dynamic pressure sensor; the turning point of the axial to radial energy ratio is analyzed using a vibration accelerometer; and the temperature entropy value of the inlet guide vane area is calculated based on infrared thermal imager data: , where S T is the temperature entropy, p i The distribution probability of each temperature range is represented. The fusion system needs to trigger a surge warning when the temperature entropy sudden change growth rate exceeds 0.8 bits / s and the pressure pulsation amplitude exceeds 25% of the rated value. At the same time, it needs to verify that the multi-sensor time alignment error is less than 100 μs.

[0044] S2: Data stream processing and real-time analysis: Dynamic benchmarks are generated through optical and material labeling technology to replace traditional manual calibration. Data flows through a spatiotemporal alignment pipeline to unify the sampling rate and coordinate system. Online residual analysis calculates the deviation between multi-source signals and the physical model. Frequency domain coherence monitoring verifies the correlation strength between sensors.

[0045] The data stream processing of S2 is based on the raw signal acquisition of the sensor array. All data is time-stamped based on the IEEE1588 precision time protocol. The raw signal is first temporarily stored in a circular buffer with a buffer depth set to twice the maximum expected delay. The resampling engine then performs spatiotemporal normalization: the sampling rate of each sensor is unified to a 100kHz benchmark using a cubic spline interpolation algorithm, while compensating for the microsecond delay caused by the difference in transmission cable length. During the coordinate system conversion phase, the laser tracker scans the key reference points of the device every 30 seconds to construct the spatial transformation matrix in real time: , where Δx, Δy, Δz are the position offsets, and θ is the installation deflection angle. The converted data forms a spatiotemporal synchronization cube and is input into the fusion core for multi-dimensional analysis.

[0046] The core of the real-time analysis in S2 is the residual monitoring engine, which calculates the multi-source residual matrix every 200 milliseconds: , i represents the sensor number, j corresponds to the physical model type, represents the real-time measurement value of the i-th sensor, represents the predicted value of the jth type of physical model. The vibration residual analysis uses the Campbell diagram comparison technology to match the measured acceleration spectrum with the critical speed trajectory output by the rotor dynamics model; the temperature residual is obtained by solving the three-dimensional heat conduction equation Generate theoretical temperature field, k is thermal conductivity, T is temperature, ρ is density, c p is the specific heat capacity, t is the time, is the gradient operator, When the residual kurtosis value exceeds 3.0 or five consecutive sampling points exceed the ±3σ boundary, the anomaly mark is triggered in real time and the root cause analysis thread is started.

[0047] The S2 uses frequency domain coherence monitoring as an auxiliary verification method to track the transfer characteristics of current and vibration: , is the cross power spectrum density of current and vibration, and They are respectively the autopower spectrum. In the characteristic frequency band, when the coherence coefficient is lower than 0.7 and lasts for three seconds, it is determined that the sensor is inaccurate.

[0048] The processing process adopts a pipeline architecture to ensure real-time performance: the first-level FPGA completes timestamp alignment and filtering, the second-level GPU cluster performs model calculation and residual analysis, and the third-level CPU implements logical decision-making.

[0049] S3: Multi-level logical judgment and calibration execution: Implement a four-level decision rule base and dynamic confidence weight management to diagnose sensor status and trigger calibration.

[0050] The multi-level logic judgment of S3 starts with the input of raw data. First, a single sensor self-test is performed: preliminary screening is achieved by analyzing the signal baseline stability for 10 consecutive minutes with a standard deviation of less than 0.1% of the full scale, a signal-to-noise ratio of more than 60dB, and a power supply ripple of less than 5mV. Sensors that pass the self-test enter the physical consistency check stage. At this time, the system calls the preset physical model library and compares the sensor's measured value with the model's predicted value in real time: the vibration signal must meet , represents the frequency domain amplitude of the measured vibration acceleration signal, σ represents the standard deviation, where By transferring the function and electromagnetic force Calculated theoretical value, F em represents the electromagnetic excitation force, k t represents the motor torque constant, r represents the rotor equivalent radius, Represents the current harmonic component, and the temperature distribution must conform to the heat conduction equation The numerical solution boundary of k is thermal conductivity, T is temperature, ρ is density, c p is the specific heat capacity, t is the time, is the gradient operator, is the symbol of partial derivative;

[0051] When the physical consistency check fails, the system automatically activates the model parameter correction module and updates the transfer function based on the residual gradient descent method. Or the thermal conductivity k of the material, enter the multi-sensor mutual verification link through the inspection data, and cross-compare the complex coherence function of the current harmonics and vibration spectrum :like If the vibration residual exceeds the standard, it is determined to be sensor drift. And the vibration residual exceeds the standard, triggering the equipment protection mechanism;

[0052] Sensors that fail mutual verification will initiate the fourth level of adversarial testing and verification: the step disturbance is reproduced under controlled working conditions. If the fault characteristics are reproduced, the sensor is confirmed to be failed, the data is automatically isolated and a replacement instruction is triggered; if it is not reproduced, it is determined to be a false alarm and the confidence weight W of the sensor is adjusted. i (t), confidence management adopts a dynamic weighting model: , R recent is the residual compliance rate in the last five minutes, R history Based on the exponential decay function calculation of the number of historical failures, E env E is the environmental severity factor, when the temperature is >80℃ env The value is 0.8, and the sensor data with a final confidence level lower than 0.6 will be isolated;

[0053] The calibration execution phase of S3 first identifies the drift pattern: for the bias error, perform Compensation, μ residual Represents the statistical mean of the residual sequence; the gain drift is expressed as , S raw Represents the original sensor signal, and the k value is calculated by back-calculating the adjacent sensors; nonlinear distortion activates the pre-trained neural network corrector. After calibration, a 0.5Hz sine sweep signal is immediately injected to verify phase consistency and start 24-hour Allan variance monitoring. To ensure long-term stability, , τ represents the duration of the data packet, represents the arithmetic mean of the sensor output values ​​in the kth time window, It represents the arithmetic mean of the adjacent next time window. Finally, the theoretical value of the bearing reaction force is compared with the measured value through the laser Doppler vibrometer to complete the closed-loop verification.

[0054] The present invention first implements a step-by-step anti-disturbance protocol and designs a multi-stage electromechanical thermal coupling disturbance scenario consisting of four stages to stimulate potential sensor errors. The first stage is to establish basic working conditions and inject low-frequency disturbances. After running at rated speed without load, a specific voltage fluctuation is applied, focusing on monitoring the current spectrum changes and low-frequency vibration response delay. When phase lag or temperature gradient mutation is detected, the sensor gain calibration module is triggered to perform closed-loop correction. The second stage is load step and high-frequency common-mode interference. After increasing the load, a high-frequency interference signal is injected, and electromagnetic interference path identification, signal-to-noise ratio attenuation quantification and oil contamination working condition verification are performed simultaneously. If the current measurement error exceeds the standard, the anti-aliasing filter reconstruction algorithm is activated and dual-path adaptive filtering is used to eliminate spectrum aliasing. The third stage is transient impact and dynamic stiffness modulation. Under the condition of maintaining load, current spikes and mechanical impact are injected, the overload recovery characteristics of the sensor are captured, and the natural frequency of the system is changed through hydraulic release. When the transfer function changes beyond the limit, the installation coupling failure is determined and the recalibration process is initiated. This process is completed by establishing a reference grid with a laser tracker and combining robotic arm positioning with high-speed camera dynamic tracking. The fourth stage is surge limit testing. Surge conditions are created by periodically adjusting valves, and data fusion from dynamic pressure sensors, vibration accelerometers, and infrared thermal imagers is used to trigger an early warning. Data stream processing and real-time analysis are then performed. The raw signal timestamps are aligned based on the precision time protocol and temporarily stored in a circular buffer. A resampling engine unifies the sampling rate and compensates for transmission delays. Spatial and temporal synchronization is achieved by combining the spatial transformation of the laser tracker. After input into the fusion core, the residual monitoring engine periodically calculates the multi-source residual matrix for anomaly detection. Frequency domain coherence monitoring is used to verify sensor correlation strength. Finally, multi-level logic judgment and calibration execution are implemented, sequentially executing four levels of decision-making: single sensor self-test, physical consistency check, multi-sensor mutual verification, and adversarial testing. Low-confidence sensor data is isolated through dynamic confidence weight management. During the calibration execution phase, drift patterns are identified and offset compensation, gain correction, or neural network correction is performed. After calibration, phase consistency is verified using a sine frequency sweep signal, and long-term stability monitoring is initiated. Finally, a laser Doppler vibrometer is used to compare theoretical and measured values ​​for closed-loop verification.

[0055] Secondly: The drawings of the embodiments disclosed in the present invention only involve structures related to the embodiments disclosed in the present invention. Other structures may refer to conventional designs. The same embodiment and different embodiments of the present invention may be combined with each other without conflict.

[0056] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. An experimental calibration method based on multi-sensor signal fusion processing, characterized in that: include: S1: Step-by-step counter-disturbance protocol: Design a multi-stage electromechanical and thermal coupling disturbance scenario to stimulate potential sensor errors; S2: Data stream processing and real-time analysis: Dynamic benchmarks are generated through optical and material labeling technology to replace traditional manual calibration. Data flows through a spatiotemporal alignment pipeline to unify sampling rates and coordinate systems. Online residual analysis calculates the deviation between multi-source signals and the physical model. Frequency domain coherence monitoring verifies the correlation strength between sensors. S3: Multi-level logical judgment and calibration execution: Implement a four-level decision rule base and dynamic confidence weight management to diagnose sensor status and trigger calibration; The multi-level logic judgment of S3 starts with the input of raw data. First, a single sensor self-test is performed: preliminary screening is achieved by analyzing the signal baseline stability for 10 consecutive minutes with a standard deviation of less than 0.1% of the full scale, a signal-to-noise ratio of more than 60dB, and a power supply ripple of less than 5mV. Sensors that pass the self-test enter the physical consistency check stage. At this time, the system calls the preset physical model library and compares the sensor's measured value with the model's predicted value in real time: the vibration signal must meet , represents the frequency domain amplitude of the measured vibration acceleration signal, σ represents the standard deviation, where By transferring the function and electromagnetic force Calculated theoretical value, F em represents the electromagnetic excitation force, k t represents the motor torque constant, r represents the rotor equivalent radius, Represents the current harmonic component, and the temperature distribution must conform to the heat conduction equation The numerical solution boundary of k is thermal conductivity, T is temperature, ρ is density, c p is the specific heat capacity, t is the time, is the gradient operator, is the symbol of partial derivative; When the physical consistency check fails, the system automatically activates the model parameter correction module and updates the transfer function based on the residual gradient descent method. Or the thermal conductivity k of the material, enter the multi-sensor mutual verification link through the inspection data, and cross-compare the complex coherence function of the current harmonics and vibration spectrum :like If the vibration residual exceeds the standard, it is determined to be sensor drift. And the vibration residual exceeds the standard, triggering the equipment protection mechanism; Sensors that fail mutual verification will initiate the fourth level of adversarial testing and verification: the step disturbance is reproduced under controlled working conditions. If the fault characteristics are reproduced, the sensor is confirmed to be failed, the data is automatically isolated and a replacement instruction is triggered; if it is not reproduced, it is determined to be a false alarm and the confidence weight W of the sensor is adjusted. i (t), confidence management adopts a dynamic weighting model: , R recent is the residual compliance rate in the last five minutes, R history Based on the exponential decay function calculation of the number of historical failures, E env E is the environmental severity factor, when the temperature is >80℃ env The value is 0.8, and the sensor data with a final confidence level lower than 0.6 will be isolated.

2. The experimental calibration method based on multi-sensor signal fusion processing according to claim 1, characterized in that: The first stage of the multi-stage electromechanical thermal coupling disturbance scenario in S1 is the establishment of basic working conditions and low-frequency disturbance injection; the second stage is load step and high-frequency common-mode interference; the third stage is transient impact and dynamic stiffness modulation; and the fourth stage is surge boundary limit test.

3. The experimental calibration method based on multi-sensor signal fusion processing according to claim 2, characterized in that: The basic working condition establishment and low-frequency disturbance injection first maintain the rated speed for ten minutes in the no-load state of the equipment, collect sensor baseline data of vibration, temperature, current and oil parameters, and then inject the first-level disturbance: superimpose ±5% 0.5 times the power frequency voltage fluctuation on the power input side through the programmable frequency converter. The fluctuation frequency is 25Hz and lasts for five minutes. The first stage focuses on monitoring the harmonic spectrum changes of the current transformer and the response delay of the low-frequency component of the bearing seat vibration, where the low-frequency component is the frequency band below 200Hz, and synchronously records the temperature distribution uniformity of the stator core surface with an infrared thermal imager. When the vibration phase lag exceeds 15° or the temperature change of the stator core surface exceeds 8°C per meter along the axial direction is detected, the sensor gain calibration module is triggered.

4. The experimental calibration method based on multi-sensor signal fusion processing according to claim 2, characterized in that: The load step and high-frequency common-mode interference increase the equipment load step to 80% of the rated value. After the operation stabilizes, the second-level disturbance is injected: a high-frequency common-mode interference signal with a frequency of 5kHz and a peak-to-peak value of 50Vpp is injected into the control loop for 120 seconds. In the second stage, three operations are performed simultaneously: a high-voltage differential probe is used to capture the transient voltage waveform of the motor terminal to identify the electromagnetic interference intrusion path; a laser Doppler vibrometer is turned on to scan the rotor journal area to quantify the signal-to-noise ratio attenuation curve under high-frequency interference; ISO 4406 grade 16 standard particles with a particle size distribution range of 15-25μm are injected into the lubrication oil circuit, and the fidelity of the pollution condition signal is verified by correlation analysis between electrostatic sensors and high-speed microscopy. When the measurement error of the current harmonic sensor at the 5kHz frequency point exceeds ±3%, the anti-aliasing filter reconstruction algorithm is automatically activated.

5. The experimental calibration method based on multi-sensor signal fusion processing according to claim 2, characterized in that: The transient shock and dynamic stiffness modulation injects a third-level disturbance while maintaining an 80% load condition: a current spike with a peak value of 200A and a pulse width of 100μs is generated through a preset short-circuit device, and a pneumatic exciter is synchronously triggered to apply a 3ms half-sine mechanical shock with an acceleration amplitude of 15 times the acceleration of gravity. A high-speed acquisition card with a sampling rate of 100MS / s is used to capture the sensor overload recovery characteristics. The time required for the signal to recover to the normal fluctuation band is no more than 50ms, where the normal fluctuation band is defined as the range of ±2 times the standard deviation. At the same time, 50% of the preload of the base is hydraulically released, causing the natural frequency to drop from 500Hz to 200Hz. The hydraulic release of the base preload takes 3 minutes. The foundation displacement of the equipment is monitored in real time by a laser tracker to construct the vibration transfer function matrix: ,in represents the new transfer function, Vib bearing Indicates the bearing vibration spectrum, Vib base Represents the basic vibration spectrum, J is the imaginary unit, which is the primitive of the negative domain, and the physical meaning of -J is phase lag. is the phase angle. When the transfer function amplitude changes by more than 20 dB or the phase shift exceeds 30°, the sensor installation coupling is determined to be invalid and the recalibration process is started.

6. The experimental calibration method based on multi-sensor signal fusion processing according to claim 2, characterized in that: The surge boundary limit test closes the outlet valve to a critical opening range of 65%-70% to create a progressive surge condition: the pressure relief valve is periodically adjusted at a frequency of 0.2 Hz, with its opening changing in steps between 30% and 80%. The propagation velocity of the flow channel pulsation wavefront is captured using a dynamic pressure sensor; the turning point of the axial to radial energy ratio is analyzed using a vibration accelerometer; and the temperature entropy value of the inlet guide vane area is calculated based on infrared thermal imager data: , where S T is the temperature entropy, p i The distribution probability of each temperature range is represented. The fusion system needs to trigger a surge warning when the temperature entropy sudden change growth rate exceeds 0.8 bits / s and the pressure pulsation amplitude exceeds 25% of the rated value. At the same time, it needs to verify that the multi-sensor time alignment error is less than 100 μs.

7. The experimental calibration method based on multi-sensor signal fusion processing according to claim 1, characterized in that: The S2 data stream processing is based on the raw signal acquisition of the sensor array. All data is time-stamped based on the IEEE 1588 precision time protocol. The raw signal is first temporarily stored in a circular buffer with a buffer depth set to twice the maximum expected delay. The resampling engine then performs spatiotemporal normalization: the sampling rate of each sensor is unified to a 100kHz benchmark using a cubic spline interpolation algorithm, while compensating for microsecond delays caused by differences in transmission cable lengths. During the coordinate system conversion phase, the laser tracker scans the key reference points of the device every 30 seconds, constructing a spatial transformation matrix in real time: , where Δx, Δy, Δz are the position offsets, and θ is the installation deflection angle. The converted data forms a spatiotemporal synchronization cube and is input into the fusion core for multi-dimensional analysis.

8. The experimental calibration method based on multi-sensor signal fusion processing according to claim 1, characterized in that: The core of the real-time analysis in S2 is the residual monitoring engine, which calculates the multi-source residual matrix every 200 milliseconds: , i represents the sensor number, j corresponds to the physical model type, represents the real-time measurement value of the i-th sensor, represents the predicted value of the jth type of physical model. The vibration residual analysis uses the Campbell diagram comparison technology to match the measured acceleration spectrum with the critical speed trajectory output by the rotor dynamics model; the temperature residual is obtained by solving the three-dimensional heat conduction equation Generate theoretical temperature field, k is thermal conductivity, T is temperature, ρ is density, c p is the specific heat capacity, t is the time, is the gradient operator, When the residual kurtosis value exceeds 3.0 or five consecutive sampling points exceed the ±3σ boundary, the anomaly mark is triggered in real time and the root cause analysis thread is started.

9. The experimental calibration method based on multi-sensor signal fusion processing according to claim 1, characterized in that: The calibration execution phase of S3 first identifies the drift pattern: for the bias error, perform Compensation, μ residual Represents the statistical mean of the residual sequence; the gain drift is expressed as , S raw Represents the original sensor signal, and the k value is calculated by back-calculating the adjacent sensors; nonlinear distortion activates the pre-trained neural network corrector. After calibration, a 0.5Hz sine sweep signal is immediately injected to verify phase consistency and start 24-hour Allan variance monitoring. To ensure long-term stability, , τ represents the duration of the data packet, represents the arithmetic mean of the sensor output values ​​in the kth time window, It represents the arithmetic mean of the adjacent next time window. Finally, the theoretical value of the bearing reaction force is compared with the measured value through the laser Doppler vibrometer to complete the closed-loop verification.

Images