Instantaneous torque measuring system and method based on engine front end

Abstract

The invention discloses an instantaneous torque measuring system and method based on the front end of an engine, and relates to the technical field of engine testing and signal processing, and the system comprises a signal acquisition module, a noise analysis module, a signal processing module and a signal processing module, the noise frequency distribution and noise amplitude range extraction module is used for extracting noise frequency distribution and a noise amplitude range from the smoothed torque data and determining filtered torque data containing low-frequency drift noise and periodic interference signals, and the parameter extraction module is used for analyzing the noise frequency distribution and the periodic interference signals by adopting fast Fourier transform according to the filtered torque data so as to obtain the low-frequency drift noise and the periodic interference signals. Acquiring noise duration time and a noise source identifier, and determining an amplitude compensation coefficient and a frequency suppression parameter; according to the instantaneous torque measuring system and method based on the engine front end, accurate acquisition and noise suppression of engine torque signals are achieved, the accuracy and reliability of torque data are improved, and effective support is provided for engine performance monitoring and fault diagnosis.

Description

Technical Field

[0001] The present invention relates to the technical field of engine testing and signal processing, and particularly relates to an instantaneous torque measurement system and method based on the front end of an engine. Background Art

[0002] In modern engine testing and control systems, torque measurement, as an important link for power performance evaluation and control strategy optimization, places extremely high requirements on the accuracy and real-time performance of the measurement system. Existing engine torque measurement systems usually rely on high-frequency sensors to obtain the original torque signal, and perform filtering and analysis through a digital signal processing module to achieve effective extraction and real-time transmission of torque data. However, when operating in a high-noise environment, such systems face various technical bottlenecks.

[0003] Firstly, due to the frequent and complex noise sources of the engine itself and its surrounding working environment, the torque signals collected by the sensor often contain a large amount of high-frequency interference components. Directly using these original signals will cause measurement data distortion, seriously affecting the system accuracy. Although traditional digital filtering methods (such as low-pass filters, Kalman filters, etc.) can suppress noise to a certain extent, these methods generally have problems of phase delay, which further destroys the timing characteristics of the measurement data and is not conducive to the real-time response of subsequent control logic. Secondly, the noise characteristics vary significantly under different engine operating conditions (such as different speeds, loads, or combustion modes), resulting in difficulty for fixed-parameter filtering algorithms to work stably under all operating conditions. Even when using adaptive filtering algorithms, there are often problems such as high computational complexity and slow convergence speed, making it difficult to balance filtering performance and processing efficiency. In addition, after signal processing, the system needs to transmit the processed high-precision torque data to the host computer or control unit. However, due to factors such as data link bandwidth, communication protocol delay, and coding method, it is difficult for the system to ensure real-time transmission of large-scale and high-frequency sampled data, seriously restricting the application of data in scenarios such as high-speed testing and closed-loop control. Summary of the Invention

[0004] The purpose of the present invention is to provide an instantaneous torque measurement system and method based on the front end of an engine to solve the problems existing in the prior art.

[0005] To achieve the above purpose, the present invention provides the following technical solution: An instantaneous torque measurement system based on the front end of an engine, comprising: A signal acquisition module, configured to acquire the torque signal at the front end of the engine and perform smoothing processing; A noise analysis module, configured to extract the noise frequency distribution and noise amplitude range from the smoothed torque data, and determine the filtered torque data including low-frequency drift noise and periodic interference signals; A parameter extraction module, which is used to analyze the noise frequency distribution and periodic interference signals of the filtered torque data by using fast Fourier transform, obtain the noise duration and noise source identification, and determine the amplitude compensation coefficient and frequency suppression parameter; A data correction module, which is used to adjust the filtered torque data by the amplitude compensation coefficient and drift correction offset if the noise source identification indicates that the noise comes from the sensor; if the noise source identification indicates a mechanical component, it is corrected by the frequency suppression parameter and random noise attenuation to generate the corrected torque data; A transmission module, which is used to segment the corrected torque data according to the compensation time window through a real-time data transmission protocol, perform format conversion with a dynamically adjusted step size, generate data packets that conform to the target device protocol, and limit the adjustment range of the data packets not to exceed the compensation saturation threshold; A target device parsing module, which is used to parse the data packets by the target device, obtain the corrected torque data, verify whether the noise amplitude range is lower than the preset threshold, and determine the final torque data.

[0006] Preferably, the signal acquisition module collects the engine front-end torque signal through a high-frequency sensor, and uses a differential amplifier circuit to amplify the torque signal to obtain the amplified torque signal.

[0007] Preferably, the signal acquisition module also performs analog-to-digital conversion on the amplified torque signal to generate a digital torque signal, and extracts the initial torque data containing high-frequency noise components and low-frequency drift noise according to the digital torque signal.

[0008] Preferably, the signal acquisition module also performs digital processing on the initial torque data according to the preset sampling frequency to generate instantaneous torque data containing periodic interference signals and random white noise.

[0009] Preferably, the signal acquisition module also determines that the type of the digital filter is a low-pass filter according to the preset filtering parameters, and performs signal smoothing processing on the high-frequency noise components in the instantaneous torque data by using the low-pass filter to obtain the smoothed torque data.

[0010] Preferably, the noise analysis module is further used to construct a noise power spectral density map of the torque signal to distinguish low-frequency drift and periodic interference signals.

[0011] Preferably, the parameter extraction module uses the fast Fourier transform method optimized by a window function to reduce spectral leakage and improve the frequency extraction accuracy.

[0012] Preferably, the data correction module includes a group of noise source identification models, in which the noise source is discriminated based on historical working condition characteristics and sensor types to dynamically select different correction paths.

[0013] Preferably, the target device parsing module includes a decoder, a reconstruction logic unit, and an error checking mechanism, which are used to check data integrity after packet decoding and trace back abnormal data segments.

[0014] An instantaneous torque measurement method based on the front end of an engine, using the instantaneous torque measurement system based on the front end of the engine, the method includes: Collect the torque signal at the front end of the engine and perform smoothing processing; Extract the noise frequency distribution and noise amplitude range from the smoothed torque data, and determine the filtered torque data including low-frequency drift noise and periodic interference signals; For the filtered torque data, use fast Fourier transform to analyze the noise frequency distribution and periodic interference signals, obtain the noise duration and noise source identifier, and determine the amplitude compensation coefficient and frequency suppression parameter; If the noise source identifier indicates that the noise comes from the sensor, adjust the filtered torque data through the amplitude compensation coefficient and drift correction offset; if the noise source identifier indicates a mechanical component, correct it through the frequency suppression parameter and random noise attenuation to generate the corrected torque data; Through the real-time data transmission protocol, segment the corrected torque data according to the compensation time window, perform format conversion with a dynamically adjusted step size, generate data packets that conform to the target device protocol, and limit the adjustment range of the data packets not to exceed the compensation saturation threshold; Parse the data packet through the target device, obtain the corrected torque data, verify whether the noise amplitude range is lower than the preset threshold, and determine the final torque data.

[0015] It can be seen from the above technical solutions that the present invention has the following beneficial effects: The instantaneous torque measurement system and method based on the front end of the engine collect the torque signal at the front end of the engine through a high-frequency sensor and perform amplification and digital processing to obtain the initial torque data containing high-frequency noise and low-frequency drift. For different types of noise, the present invention uses a low-pass filter for signal smoothing and analyzes the noise characteristics through fast Fourier transform to determine the noise source. According to the noise source, the present invention uses methods such as amplitude compensation or frequency suppression for correction to generate the corrected torque data. Finally, the present invention segments the corrected data according to the time window, performs format conversion and generates data packets that conform to the target device protocol, realizing the accurate acquisition and noise suppression of the engine torque signal, improving the accuracy and reliability of the torque data, and providing effective support for engine performance monitoring and fault diagnosis. Description of the Drawings

[0016] Figure 1 It is a connection diagram of the system modules of the present invention; Figure 2 This is the flowchart of the method of the present invention. Specific embodiments

[0017] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0018] As Figure 1 shown, the present invention provides a technical solution: an instantaneous torque measurement system based on the front end of an engine, including: A signal acquisition module, configured to acquire the torque signal at the front end of the engine and perform smoothing processing; A noise analysis module, configured to extract the noise frequency distribution and noise amplitude range from the smoothed torque data, and determine the filtered torque data including low-frequency drift noise and periodic interference signals; A parameter extraction module, configured to perform fast Fourier transform analysis on the filtered torque data to analyze the noise frequency distribution and periodic interference signals, obtain the noise duration and noise source identifier, and determine the amplitude compensation coefficient and frequency suppression parameter; A data correction module, configured to, if the noise source identifier indicates that the noise comes from the sensor, adjust the filtered torque data through the amplitude compensation coefficient and drift correction offset; if the noise source identifier indicates a mechanical component, perform correction through the frequency suppression parameter and random noise attenuation to generate the corrected torque data; A transmission module, configured to segment the corrected torque data according to the compensation time window through a real-time data transmission protocol, perform format conversion with a dynamically adjusted step size, generate a data packet conforming to the protocol of the target device, and limit the adjustment amplitude of the data packet not to exceed the compensation saturation threshold; A target device parsing module, configured to parse the data packet by the target device, obtain the corrected torque data, verify whether the noise amplitude range is lower than a preset threshold, and determine the final torque data.

[0019] The system performs full - process noise stripping and correction based on the torque signal at the front end of the engine. First, the signal acquisition module obtains the original torque data, and a weighted moving average algorithm is used for preliminary smoothing to suppress instantaneous peak values. Then, the noise analysis module uses wavelet transform to extract the frequency - domain features in the torque data and identify low - frequency drift and periodic noise. The parameter extraction module uses the fast Fourier transform (FFT) to perform refined frequency decomposition on these noise features, thereby analyzing the frequency range and duration of the noise, and determining the noise source in combination with the pattern recognition algorithm. According to different noise sources, the system calls the data correction module to perform targeted correction on the filtered data through amplitude compensation, offset correction, or frequency attenuation processing methods. Finally, the torque data is converted into a format that conforms to the real - time bus protocol (such as CAN, EtherCAT, etc.) through the transmission module, and it is ensured that the compensation amplitude does not exceed the set saturation threshold. The target device then parses the data packet to verify its stability and accuracy.

[0020] The key of this system lies in the real - time acquisition, noise feature analysis, parameter identification, and compensation correction of the instantaneous torque signal at the front end of the engine. The system first obtains the original torque signal in the form of a time series through the signal acquisition module , and performs smoothing processing using the weighted moving average algorithm: ; where, represents the smoothed torque signal value, represents the original torque signal, t−i represents the offset of this original data point at the time point, represents the weight of the i - th time - delay window, and the weight can be set according to exponential decay. t represents the current time index, and N is the window width (typical value is 10 - 20).

[0021] The noise analysis module uses wavelet transform to analyze the frequency features in the smoothed signal : ; where, is the wavelet basis function, is the scale parameter, is the translation parameter, is the normalization factor. By extracting the region with the largest modulus value of the wavelet coefficients, low - frequency drift (typical frequency < 1Hz) and high - frequency interference (such as engine ignition periodic interference in the range of dozens of Hz) can be identified.

[0022] The parameter extraction module performs the fast Fourier transform on the filtered signal: ; where, represents the complex amplitude of the torque signal at the k-th frequency, including the energy (magnitude) and phase information of this frequency component. k is the frequency index. represents the value of the instantaneous engine torque signal after filtering. n represents the n-th sampling time point, N represents the total number of sampling points, and j is used to describe the phase component of the signal. represents the Fourier basis function.

[0023] Obtain the spectral distribution and extract the main interference frequencies by combining the peak recognition algorithm. and the amplitude . Thus, determine the noise duration. and judge whether the noise source is the sensor (electronic noise) or mechanical components (resonance / wear) through the signal source library comparison model.

[0024] If it is sensor noise, apply amplitude compensation: ; Wherein, is the compensated torque signal, is the filtered torque signal, is the amplitude compensation coefficient (generally set to 0.6 - 0.9 according to experience), is the measured average deviation value.

[0025] If it is mechanical interference, then adopt the frequency suppression method: ; Wherein, is the torque signal after noise correction, is the filtered torque signal, is the suppression amplitude, is the interference frequency, is the phase compensation term, and all parameters are dynamically generated by frequency domain analysis. is the main interference frequency, represents the sine wave component corresponding to the noise frequency and represents the time-domain signal corresponding to this noise frequency.

[0026] The transmission module segments according to the time window and adjusts the step size according to the transmission bandwidth, and converts it into the data packet format, using the following formatting function: ; Wherein, is the packed torque data segment, Format() is the formatting function, which encapsulates the input signal and its parameters into a standard structure. is the window time length, is the current step adjustment coefficient, and it is restricted that: ; is the compensation saturation threshold to prevent mutation.

[0027] After receiving the data packet, the target device parsing module decodes it and performs noise amplitude verification: ; Among them, is the noise amplitude, is the maximum residual noise tolerance set by the system (such as ±0.05 Nm). If the condition is met, the final torque data is output .

[0028] This system provides an automated and high-precision instantaneous torque measurement solution for the whole process from torque signal acquisition to data packet parsing. Through the dual-frequency feature recognition mechanism of wavelet analysis and Fourier transform, complex interference sources are effectively identified. The multi-strategy compensation and correction model is adopted to improve the data recovery accuracy and robustness. The data transmission part combines the real-time protocol and the saturation control mechanism to enhance the data adaptability of the system under high-dynamic working conditions and avoid transmission distortion. Finally, the engineering usability and data credibility of the measurement system are improved.

[0029] In the power test center of a diesel engine manufacturing enterprise, in order to improve the real-time performance and accuracy of engine bench test data, the technical team introduced the instantaneous torque measurement system based on the engine front end described in this application. This system is applied to the power assembly test process of a 4.5L in-line four-cylinder turbocharged diesel engine. The main goal is to capture the torque dynamic response data from startup to acceleration and identify the periodic interference sources among them.

[0030] An integrated torque sensor (model TQ-PT450) is installed in the flywheel housing area at the engine front end and connected to the data acquisition module of this system through the CAN interface. The acquisition module is embedded with a filtering algorithm and connected to the analysis and processing platform deployed in the industrial computer. After the engine starts, the system continuously acquires torque signals at a sampling frequency of 1 kHz , the sliding average algorithm with exponential weights is adopted to preliminarily smooth the signal and eliminate spikes. Through wavelet transform, a group of low-frequency drift components (about 0.3 Hz) and high-frequency interference components (corresponding to the engine ignition frequency of 42 Hz) are identified, and their sources are judged to be the zero drift of the sensor and the structural resonance in the cylinder compression stage respectively. The fast Fourier transform (FFT) is used to extract the above frequency components, and through template matching and comparison with the known equipment database, the system automatically identifies the types of interference sources, which respectively point to "sensor electronic drift" and "structural periodic interference of the flywheel and crankshaft assembly". For the sensor error part, the system calculates the amplitude compensation coefficient α = 0.82 and superimposes the average offset for correction; for the mechanical resonance part, the main frequency amplitude is extracted and the frequency suppression parameter is set to complete the superimposed attenuation based on the sine cancellation model. The corrected data is sent to the host computer platform via the EtherCAT real-time protocol. Each time window is set to 50 ms, and the step size is dynamically adjusted between 0.1 and 0.2 ms. The system automatically controls the adjustment amplitude of each packet of data not to exceed 0.05 Nm. After receiving the data packet, the target device performs the verification of the noise amplitude range. All the resolved residual noise is controlled within ±0.04 Nm, meeting the engineering specification of the test center for the control closed-loop feedback data error to be less than ±0.05 Nm. This system shows good robustness and consistency in 5 repeated tests. The torque change curve in the startup stage is smoother than that of the traditional system, which can accurately capture the rapid change trend and provide reliable data support for subsequent fuel injection calibration and vibration and noise control strategy development.

[0031] The signal acquisition module collects the engine front-end torque signal through a high-frequency sensor, and uses a differential amplifier circuit to amplify the torque signal to obtain the amplified torque signal. The key point of this implementation is the optimization of the enhanced acquisition ability and anti-interference performance of the signal acquisition module. First, the high-frequency sensor is installed inside the engine flywheel housing, and the original electrical signal is generated by measuring the small angular displacement or stress change of the torque transmission shaft. , such sensors have the characteristics of high response frequency (the bandwidth range can reach hundreds of Hz) and strong anti-environmental disturbance ability, and are suitable for capturing torque changes under high-dynamic working conditions. Since the original signal level is low (usually in the millivolt level) and is easily affected by environmental electromagnetic interference (EMI) and ground potential difference interference, the system performs anti-interference amplification processing through a differential amplifier circuit. The structure of the differential amplifier is as follows: ; Among them, is the output voltage of the amplifier, are the positive and negative input signals output by the high-frequency sensor; is the differential mode gain, usually set to 50 - 100 times; is the common-mode gain, designed to be close to 0 to maximize the suppression of common-mode interference.

[0032] The system actually adopts an instrumentation amplifier architecture (such as INA333 or AD8421 chips), sets the input impedance higher than 1 MΩ to ensure good matching with the sensor, and the amplitude of the amplified signal is stable at 0 - 5V, meeting the sampling input requirements of the subsequent ADC module.

[0033] By adopting a sensor with high-frequency response combined with a differential amplification circuit processing method, the quality of torque signal acquisition of the system is greatly improved. On the one hand, the differential amplification structure effectively suppresses common-mode noise and ground loop interference, improves the signal-to-noise ratio, and ensures that the torque detail signal is not submerged by interference; on the other hand, the high-frequency sensor has higher response accuracy for instantaneous rapid changes, enabling the system to accurately reflect the torque changes in dynamic processes such as rapid start-stop and sudden load changes, and improving the overall sensitivity and response speed of the system.

[0034] The signal acquisition module also performs analog-to-digital conversion on the amplified torque signal to generate a digital torque signal, and extracts the initial torque data containing high-frequency noise components and low-frequency drift noise according to the digital torque signal.

[0035] In this embodiment, the signal acquisition module further enhances the digital processing ability of the torque signal. First, the amplified analog torque signal is subjected to real-time digital processing by a high-precision analog-to-digital converter (ADC). The ADC sampling frequency is usually set above 2 kHz to ensure sufficient sampling of high-frequency disturbances.

[0036] The analog-to-digital conversion process can be expressed as: ; where is the sampling period; ADC() is the analog-to-digital conversion function; is the current sampling point; is the discrete torque data, representing the result after ADC, is the analog signal after being processed by the amplifier.

[0037] The converted digital signal is sent to the preprocessing unit for preliminary noise analysis. The system uses a band-pass filter (BPF) to separate the effective torque information from the background interference, and at the same time extracts two types of typical noises: High-frequency noise components: originating from internal electronic fluctuations of the sensor, electromagnetic interference, etc., whose frequency is usually above several hundred Hz; Low-frequency drift noise: manifested as a slow fluctuation of the torque signal baseline, mainly caused by temperature drift, zero-point offset, etc., with a frequency lower than 1 Hz.

[0038] To identify these components, the system calculates the spectral density function of the digital signal and sets a frequency threshold ; ; is the power spectral density of the signal in the frequency domain, is a complex exponential function representing the rotation factor in the frequency domain, f is the frequency, N is the length of the discrete-time signal, n is the time index, and according to the thresholds and , the system extracts the energies of the low-frequency drift band and the high-frequency noise band respectively and marks them as the noise components to be processed in the initial torque data: (noisy data) This initial torque data will then enter the subsequent filtering and compensation processes.

[0039] In this embodiment, analog-to-digital conversion is performed immediately after the analog signal is amplified, effectively avoiding the problem of interference accumulation that may be caused by long-distance transmission of analog signals. At the same time, through frequency-domain identification technology, the system can quickly locate and mark the high-frequency and low-frequency noise regions, providing data basis for subsequent filtering algorithms (such as band-stop filtering, drift suppression), and improving the accuracy and efficiency of noise stripping. In addition, the digitalized signal is convenient for storage and reuse, which helps with later data modeling and algorithm iteration.

[0040] The signal acquisition module also digitally processes the initial torque data according to the preset sampling frequency to generate instantaneous torque data containing periodic interference signals and random white noise.

[0041] The key of this embodiment lies in digitally modeling and identifying the original torque signal containing various interference types by setting reasonable sampling frequencies and digital processing strategies. Based on the initial torque data generated in the previous embodiment, the system executes the standard digital signal processing process according to the preset sampling frequency (e.g., 2 kHz) to perform secondary refinement processing on the signal and extract periodic interference and random noise.

[0042] Periodic interference mainly comes from mechanical structures (such as crankshaft imbalance, ignition cycle fluctuations) and can be identified through Fourier transform; while random white noise is mainly composed of sensor thermal noise and line interference, showing a broadband, high-frequency, and small-amplitude distribution.

[0043] The signal model can be expressed as: ; Where: is the instantaneous torque data after digital processing; is the true torque signal; is the periodic interference signal, usually satisfying: ; Wherein, is the amplitude of the k-th frequency component; fk is the k-th disturbance frequency; is the sampling period; is the initial phase, K represents the total number of frequency components, k represents the number of the frequency component currently being summed, and n is the discrete time index; is white noise, satisfying zero mean and Gaussian distribution: ; Wherein, E[] is the expected value, Var[] is the variance, is the noise variance; The system identifies each component through sliding window FFT and covariance analysis, and estimates its amplitude and phase ; At the same time, calculate the estimated value of the white noise energy density to evaluate the current signal contamination degree and guide the setting of subsequent filtering weights.

[0044] This processing mechanism realizes the dual recognition ability of structural interference and random perturbation in torque data, and establishes a "noise model" closer to the actual operating state. Digital processing not only enhances the signal visualization, analysis and storage capabilities, but also provides accurate parameters for subsequent frequency suppression and dynamic filtering. With the help of this mechanism, the system can automatically adapt to the interference characteristics under different working conditions, and improve the overall robustness and precision control ability.

[0045] The signal acquisition module also determines that the type of the digital filter is a low-pass filter according to the preset filtering parameters, and performs signal smoothing processing on the high-frequency noise components in the instantaneous torque data by using the low-pass filter to obtain the smoothed torque data.

[0046] This implementation further clarifies the elimination strategy for high-frequency noise components in the digital processing stage, and uses a low-pass filter to perform signal smoothing processing on the torque data to remove the invalid perturbations with frequencies higher than the target torque frequency band. The system first identifies the noise frequency band from the instantaneous torque data . Combining the previous frequency domain analysis results, set the filtering parameters, including the cut-off frequency and the filter order and the window function type (such as Hamming window), etc.

[0047] When selecting a FIR type low-pass filter, its impulse response is of finite length: ; Among them, are low-pass filter coefficients, designed by the window function design method or the least mean square error method; is the filter length; is the torque data after smoothing, represents the original signal at the time index n−k.

[0048] If an IIR low-pass filter is adopted, such as a Butterworth or Chebyshev filter, the system uses the following difference equation: ; Among them, is the smoothed signal, is the instantaneous signal are filter coefficients, N represents the order of the filter, and are obtained through design tools (such as MATLAB filter designer) or numerical optimization methods.

[0049] The system usually sets the cut-off frequency to about 1.2 times the upper limit of the target signal frequency to ensure retaining effective information while removing high-frequency white noise interference. The filtered signal not only has a more continuous waveform but also enhances the accuracy of subsequent Fourier analysis and model construction.

[0050] This implementation significantly improves the smoothness and stability of the torque data, eliminates the interference of high-frequency noise on the instantaneous measurement value, and improves the overall quality of signal processing. Compared with the unfiltered data, the smoothed data has higher response controllability in the control system, effectively reducing the risk of false triggering and misjudgment. It is particularly suitable for control scenarios where the data changes violently under high-speed working conditions but does not require high-frequency accuracy.

[0051] The noise analysis module is further used to construct the noise power spectral density map of the torque signal to distinguish low-frequency drift and periodic interference signals.

[0052] This implementation further improves the analytical ability of the noise analysis module for the signal noise structure. The system is based on the digital torque signal sequence and calculates its power spectral density (Power Spectral Density, PSD) to obtain the energy distribution characteristics of the signal in the frequency domain. The calculation of PSD adopts the Welch method, and its basic steps are as follows: Segment the torque signal into several overlapping windows, each window has a length of L, and the overlapping ratio is generally 50%; Apply a window function (such as Hamming window) to each segment of the signal and perform a fast Fourier transform; Calculate the energy of each segment of the spectrum and take the average to finally form a smooth PSD estimation graph.

[0053] The mathematical expression is: ; where is the signal of the k-th segment, is the window function, is the total number of segments, is the averaging factor, is the frequency at which the power spectral density value is located, and FFT represents the discrete Fourier transform.

[0054] By plotting the PSD graph, the system can clearly distinguish the following two types of frequency characteristics: Low-frequency drift (<1 Hz): Usually presented as a high-amplitude region near the starting point of the frequency axis, with a slow upward trend; Periodic interference signals (usually in the range of dozens of Hz): Appear as clear narrow-band spikes in the PSD graph, corresponding to the periodic mechanical actions of the engine.

[0055] This module divides the frequency space into a low-frequency band and an interference band and marks the energy distribution regions respectively for subsequent noise classification and targeted compensation.

[0056] By constructing the PSD graph, the system can achieve a more refined and visual identification of the noise sources of the torque signal, overcoming the limitation of difficult separation of multi-frequency interference signals in the time domain. This frequency-domain processing method improves the resolution ability for complex composite noises (such as structural resonance superimposed with low-frequency temperature drift), provides strong support for subsequent filter parameter setting and noise type identification. At the same time, this method improves the intelligent analysis ability of the system and facilitates the adaptive optimization of the algorithm.

[0057] The parameter extraction module adopts the fast Fourier transform method optimized by the window function to reduce spectral leakage and improve the frequency extraction accuracy.

[0058] In this embodiment, the parameter extraction module enhances the frequency-domain analysis accuracy of the FFT (fast Fourier transform) in processing the torque signal with a finite length by introducing the window function optimization technology. Since the sampled signal in reality has a time-window limit, directly using the FFT for processing will cause the "spectral leakage" phenomenon, affecting the frequency extraction accuracy. Therefore, the system multiplies the input signal by the window function before the FFT operation to control the main lobe width and side lobe leakage of the spectrum. The processing steps are as follows: Perform window function processing on the instantaneous torque signal with a length of : ; Among them, is the window function, is the original signal, is the signal after windowing; Perform FFT operation on the windowed signal: ; Spectrum modulus represents the amplitude of each frequency component. The system identifies the noise frequency according to the peak location method and the amplitude-frequency gradient analysis method , amplitude and phase .

[0059] Window function The selection of has a significant impact on the spectrum quality. The system defaults to using the Hamming window: ; Its main lobe is narrow and the side lobes decay rapidly, which is suitable for identifying weak periodic interference components. If it is necessary to improve the main lobe resolution, the system can also switch to window types such as Blackman and Kaiser.

[0060] The optimized spectrum information is used by the parameter extraction module to generate the following key indicators: Noise frequency list ; Noise amplitude and energy spectrum , phase parameter , duration or occupancy interval.

[0061] Provide quantitative parameter support for the subsequent correction module.

[0062] By introducing the window function processing mechanism, this system significantly reduces the spectrum leakage problem in FFT processing, effectively suppresses the interference of non-target frequency components, and improves the resolution and stability of frequency identification. This mechanism is especially suitable for torque signal scenarios where the high-frequency vibration characteristics are dense or the frequency intervals are close, laying a solid foundation for accurately identifying periodic noise and establishing an accurate spectrum model, and contributing to the fine adjustment of the subsequent suppression mechanism.

[0063] The data correction module includes a group of noise source identification models, which discriminate the noise source based on historical operating conditions and sensor types to dynamically select different correction paths.

[0064] This embodiment aims to improve the intelligent recognition and response capabilities of the data correction module for noise characteristics. The system introduces a set of multi-dimensional noise source recognition models to classify and identify different types of noise sources (such as sensor drift, mechanical structure resonance, environmental electromagnetic disturbances, etc.), and dynamically selects a matching correction path based on this. The model inputs include historical operating condition feature datasets : such as rotational speed, load, temperature, vibration acceleration; current sensor type and parameters (such as strain gauges, capacitive, magnetostrictive, etc.); real-time torque signal spectrum characteristics (such as main frequency, amplitude, duration); environmental disturbance logs (such as communication interference, power stability data, etc.).

[0065] The system uses a discriminative noise source classification model to output noise source labels The model structure can be expressed as: ; Among them, is the above-mentioned multi-dimensional feature vector; are model parameters, is the set of noise source categories, represents the conditional probability that the output label y is category c given the input feature x and model parameters θ, and argmax is the parameter for finding the maximum value.

[0066] The model can be a decision tree, random forest, SVM, or lightweight neural network, depending on the computing resources. After recognition, the system enters the following correction path selection logic: Sensor type noise: Select the correction path based on amplitude compensation and zero drift correction; Mechanical type noise: Invoke the frequency suppression and harmonic compensation algorithm; EMI type noise: Apply a high-frequency notch filter and record signal interference events; Unknown type: Trigger an alarm and retain the data for future reference, and do not correct it temporarily.

[0067] This dynamic path mechanism avoids the problem of poor adaptation of the fixed process to complex noise patterns, and improves the robustness and intelligence level of the system.

[0068] By introducing a noise source identification model and constructing an intelligent classification system based on operating conditions and sensor characteristics, this embodiment significantly improves the judgment accuracy and processing efficiency in complex and multi-source noise environments. Dynamically selecting the correction path not only avoids the problems of overcompensation or miscompensation but also enables the correction process to have pertinence and optimization capabilities, especially suitable for multi-platform and multi-sensor type deployment scenarios. In addition, this module has the ability to continuously learn and update, which is beneficial to building an intelligent data quality management system.

[0069] The target device parsing module includes a decoder, a reconstruction logic unit, and an error checking mechanism, which are used to check the data integrity after the data packet is decoded and trace back the abnormal data segment.

[0070] This embodiment focuses on enhancing the parsing, verification, and abnormal handling capabilities of data packets in the target device parsing module. This module includes three core sub-components: Decoder: Responsible for parsing the received data packets according to predefined protocol formats (such as CAN, EtherCAT, Modbus) , extracting information such as torque data fields, timestamps, check codes, etc. The decoding logic reorganizes the data content byte by byte according to the protocol frame structure and stores it in the buffer.

[0071] Reconstruction Logic Unit: When there are abnormalities such as lost frames, bit reversals, or incorrect in-frame order in the received data packets, the reconstruction unit calls the historical frame buffer and time window mechanism to try to repair the data content by means such as interpolation, reordering, or redundant frame reuse, and reconstruct the signal trajectory: ; where represents the value of the reconstructed signal at time , is the value of the received original data at time , Interp represents linear or high-order interpolation, and Reorder represents adjusting the order according to the timestamp.

[0072] Error Check Mechanism: Based on methods such as CRC check, frame checksum, or hash digest, perform integrity verification on each frame of data. If an error is found, the system marks the error frame number and traces the relevant frames within the nearest data window as the "abnormal segment" for subsequent analysis and traceability.

[0073] For example: ; represents the data packet Calculated cyclic redundancy check value Indicates the CRC value stored in the data packet header, which is used to compare with the calculated CRC value to ensure error-free data transmission Indicates an error flag, which is used to indicate whether a data error has occurred. If the CRC check fails (i.e., the calculated CRC does not match the CRC in the header), this flag is set to 1 to indicate an error. This module supports recording the time, frequency, and location of each error to statistically analyze data quality and provides an interface for upper-layer algorithms to decide whether to perform compensation, warning, or discard

[0074] This embodiment greatly improves the robustness and abnormal response ability of the system at the data receiving end. By combining the use of a decoder and reconstruction logic, it not only ensures the accuracy of data decoding but also allows the system to automatically recover under mild data damage conditions, improving the anti-interference performance of the system. The error check mechanism strengthens the guarantee of data integrity, supports the system to establish an abnormal monitoring mechanism, has an effective identification ability for communication anomalies and noise pseudo-codes, avoids incorrect data from interfering with the control logic, and improves the stability and security of the system

[0075] As Figure 2 shown, there is also provided an instantaneous torque measurement method based on the front end of the engine. Using the instantaneous torque measurement system based on the front end of the engine, the method includes: Collect the torque signal at the front end of the engine and perform smoothing processing Extract the noise frequency distribution and noise amplitude range from the smoothed torque data, and determine the filtered torque data including low-frequency drift noise and periodic interference signals For the filtered torque data, use fast Fourier transform to analyze the noise frequency distribution and periodic interference signals, obtain the noise duration and noise source identifier, and determine the amplitude compensation coefficient and frequency suppression parameter If the noise source identifier indicates that the noise comes from the sensor, adjust the filtered torque data through the amplitude compensation coefficient and drift correction offset; if the noise source identifier indicates a mechanical component, perform correction through the frequency suppression parameter and random noise attenuation to generate the corrected torque data Through the real-time data transmission protocol, segment the corrected torque data according to the compensation time window, perform format conversion using a dynamically adjusted step size, generate data packets that conform to the target device protocol, and limit the adjustment range of the data packets not to exceed the compensation saturation threshold Parse the data packet through the target device, obtain the corrected torque data, verify whether the noise amplitude range is lower than the preset threshold, and determine the final torque data

[0076] This method focuses on the torque data acquisition and processing process at the front end of the engine, mainly including multiple core steps such as signal acquisition, frequency-domain analysis, noise identification, compensation and correction, data packaging, and final verification. First, the system uses a torque sensor with high-frequency response to collect the original torque signal at the front end of the engine in real time. After being processed by the front-end analog circuit, the signal enters the digital processing stage. To remove sudden spikes and small-amplitude high-frequency jitters, the system smooths the original signal by using a weighted average method to process a continuous number of data points, assigning different weights to different time points, and finally generating a smoother signal with better continuity and less jitter. Next, the system performs spectral analysis on the smoothed torque data, using the fast Fourier transform method with a window function to extract the included noise frequency components and their amplitude distribution ranges. Specifically, the system divides the data into multiple time segments, weights each segment of data, and then performs frequency conversion to obtain the energy distribution at different frequency points. By analyzing this spectrum, the system can accurately identify the existence of low-frequency drift phenomena and periodic interference signals. After extracting the noise characteristics, the system further determines the source of the noise. This identification process comprehensively considers information such as frequency, amplitude, duration, historical working conditions, and sensor type, and outputs a noise type label by means of a noise source identification model. If the identification result shows that the noise mainly comes from the internal characteristics of the sensor, such as zero drift or circuit perturbation, the system will adopt an amplitude compensation method, that is, correct the data by superimposing a fixed or dynamic offset value; if the noise comes from mechanical components, such as resonance or periodic perturbation, the system will correct it by constructing sine waves corresponding to multiple interference frequencies and subtracting them from the original signal. After completing the noise compensation, the corrected torque data needs to be segmented and packaged and transmitted to the target device. During the data format conversion process, the system dynamically adjusts the transmission step size according to the current workload, but the adjustment amplitude each time is not allowed to exceed the preset maximum allowable value to avoid data mutations affecting the stability of the decoding end. At the target device end, the received data packet will first be parsed by the decoder according to the protocol to extract the torque data body. Then, the reconstruction logic in the system will check whether there are problems such as lost frames or out-of-order data. If so, it will use means such as interpolation and rearrangement to repair the error segments. Finally, the system calls an error verification mechanism, such as CRC verification, checksum calculation, or hash value comparison, to verify whether each frame of data is complete and reliable. Only the data that passes the verification will be used for subsequent control or recording, otherwise it will be marked as an abnormal data segment and traced back or discarded. In summary, this method realizes the whole-process optimization from signal acquisition to reliable data output in a modular and automated manner, has strong anti-noise ability, spectral recognition accuracy, and data integrity control ability, and is especially suitable for the engine instantaneous torque measurement scenario with significant multi-source interference and strict control requirements.

[0077] Although embodiments of the present invention have been shown and described, it will be understood by those of ordinary skill in the art that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the present invention, and the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. An instantaneous torque measurement system based on the front end of an engine, characterized in that, Including: A signal acquisition module, which is used to acquire the engine front-end torque signal and perform smoothing processing; A noise analysis module, which is used to extract the noise frequency distribution and noise amplitude range from the smoothed torque data, and determine the filtered torque data including low-frequency drift noise and periodic interference signals; A parameter extraction module, which is used to analyze the noise frequency distribution and periodic interference signals of the filtered torque data by using fast Fourier transform, obtain the noise duration and noise source identifier, and determine the amplitude compensation coefficient and frequency suppression parameter; A data correction module, which is used to adjust the filtered torque data by the amplitude compensation coefficient and drift correction offset if the noise source identifier indicates that the noise comes from the sensor; If the noise source identifier indicates a mechanical component, correction is performed through the frequency suppression parameter and random noise attenuation to generate the corrected torque data; A transmission module, which is used to segment the corrected torque data according to the compensation time window through the real-time data transmission protocol, perform format conversion with a dynamically adjusted step size, generate a data packet that conforms to the target device protocol, and limit the adjustment amplitude of the data packet not to exceed the compensation saturation threshold; A target device parsing module, which is used to parse the data packet by the target device, obtain the corrected torque data, verify whether the noise amplitude range is lower than the preset threshold, and determine the final torque data.

2. The instantaneous torque measurement system based on the front end of the engine according to claim 1, wherein: The signal acquisition module acquires the engine front-end torque signal through a high-frequency sensor, and uses a differential amplifier circuit to amplify the torque signal to obtain the amplified torque signal.

3. The instantaneous torque measurement system based on the front end of the engine according to claim 2, wherein: The signal acquisition module also performs analog-to-digital conversion on the amplified torque signal to generate a digital torque signal, and extracts the initial torque data including high-frequency noise components and low-frequency drift noise according to the digital torque signal.

4. The instantaneous torque measurement system based on the front end of the engine according to claim 3, characterized in that: The signal acquisition module also performs digital processing on the initial torque data according to the preset sampling frequency to generate instantaneous torque data including periodic interference signals and random white noise.

5. The instantaneous torque measurement system based on the front end of the engine according to claim 4, characterized in that: The signal acquisition module also determines that the type of the digital filter is a low-pass filter according to the preset filtering parameters, and performs signal smoothing processing on the high-frequency noise components in the instantaneous torque data by using the low-pass filter to obtain the smoothed torque data.

6. A transient torque measurement system based on the front end of an engine according to claim 1, characterized in that: The noise analysis module is further used to construct a noise power spectral density map of the torque signal to distinguish low-frequency drift and periodic interference signals.

7. A transient torque measurement system based on the front end of an engine according to claim 1, characterized in that: The parameter extraction module uses the fast Fourier transform method optimized by a window function to reduce spectral leakage and improve the frequency extraction accuracy.

8. An instantaneous torque measurement system based on the front end of an engine according to claim 1, characterized in that: The data correction module includes a set of noise source identification models, in which the noise source is discriminated based on historical working condition characteristics and sensor types to dynamically select different correction paths.

9. A transient torque measurement system based on the front end of an engine according to claim 1, characterized in that: The target device parsing module includes a decoder, a reconstruction logic unit and an error checking mechanism, which are used to check the data integrity after the data packet is decoded and trace back the abnormal data segment.

10. A method for measuring instantaneous torque based on the front end of an engine, using the instantaneous torque measurement system based on the front end of the engine according to any one of claims 1-9, characterized in that, The method includes: Acquiring the engine front-end torque signal and performing smoothing processing; Extracting the noise frequency distribution and noise amplitude range from the smoothed torque data, and determining the filtered torque data including low-frequency drift noise and periodic interference signals; For the filtered torque data, fast Fourier transform is used to analyze the noise frequency distribution and periodic interference signals, obtain the noise duration and noise source identification, and determine the amplitude compensation coefficient and frequency suppression parameter; If the noise source identification indicates that the noise comes from the sensor, the filtered torque data is adjusted by the amplitude compensation coefficient and drift correction offset; if the noise source identification indicates a mechanical component, correction is performed through the frequency suppression parameter and random noise attenuation to generate the corrected torque data; Through the real-time data transmission protocol, the corrected torque data is segmented according to the compensation time window, and format conversion is performed using a dynamically adjusted step size to generate a data packet that conforms to the target device protocol, restricting the adjustment range of the data packet not to exceed the compensation saturation threshold; The target device analyzes the data packet to obtain the corrected torque data, verifies whether the noise amplitude range is lower than the preset threshold, and determines the final torque data.

Images