Automatic detection method for dynamic balance parameters of automobile hub

Abstract

The invention discloses an automatic detection method for dynamic balance parameters of an automobile hub, and the method comprises the steps: employing a piezoelectric ceramic array self-sensing main shaft technology, enabling detection reference establishment and signal collection to be integrated in a main shaft body, and carrying out the fusion of a hub installation process and a detection reference establishment process; synchronous acquisition of multi-dimensional vibration signals is carried out; an original mixed signal is decomposed into modal components in a self-adaptive mode, and environmental vibration interference is eliminated; a FastICA-RLS hybrid algorithm is adopted to separate an independent source signal from the preliminarily purified signal, and non-stationary interference is tracked and compensated in real time; constructing a flexible rotor model based on modal parameter identification, and calculating an unbalance amount; self-adaptive Kalman filtering error compensation and prediction are carried out, and the amount of unbalance after compensation is output; and an optimal counterweight scheme is automatically calculated based on the compensated unbalance amount, and visual display and guidance are performed through a human-computer interface, so that the detection precision and the detection efficiency are improved.

Description

Technical Field

[0001] This invention relates to the field of wheel hub testing technology, and specifically to an automated testing method for dynamic balance parameters of automobile wheel hubs. Background Technology

[0002] Automotive wheel dynamic balancing refers to eliminating centrifugal torque caused by uneven mass distribution during high-speed wheel rotation by using counterweights, thus ensuring stable operation. The principle is that even in static equilibrium, if the center of mass is located on the axis of rotation, the presence of equal mass blocks on both sides of the axle will still create opposing centrifugal torques. Therefore, equal mass blocks in opposite directions are needed to achieve dynamic balance. The accuracy and sensitivity of key components in the equipment used for automotive wheel dynamic balancing testing, such as sensors, rollers, or swing frames, directly affect the test results.

[0003] Existing balancing machine software relies on electronic components such as data acquisition cards for stable operation. If the algorithm cannot effectively filter out environmental vibrations or workpiece shape and position deviations (such as uneven tire wear or wheel hub deformation), it will amplify measurement errors. For flexible rotors (such as high-speed rotating wheel hub assemblies), traditional rigid rotor balancing models may fail, requiring the use of pattern methods or relational coefficient methods. However, these methods have high requirements for calculation accuracy and real-time performance, and are prone to correction deviations due to improper parameter settings. In addition, if there are unforeseen imbalance sources in the wheel hub or tire before detection (such as tread bulges, missing lead blocks inside the wheel hub, or foreign objects attached to the tread), additional centrifugal inertial forces will be introduced, distorting the sensor signal. Furthermore, when workpieces are installed, centering errors, uneven tightening torque, or misaligned markings will also disrupt the balance state, causing the detection results to fail to reflect the true amount of imbalance. Summary of the Invention

[0004] The purpose of this invention is to provide an automated detection method for the dynamic balance parameters of automobile wheel hubs, so as to solve the problems mentioned in the background art.

[0005] The specific technical solution provided by this invention is as follows: An automated detection method for dynamic balance parameters of automobile wheel hubs, comprising the following operational steps: Step S1: Using piezoelectric ceramic array self-sensing spindle technology, the detection benchmark establishment and signal acquisition are integrated into the spindle body, and the hub installation process is merged with the detection benchmark establishment process.

[0006] Preferably, an infrared depth imaging sensor scans the appearance of the wheel hub under test to obtain feature images of bolt holes and center holes. These image features are matched with a wheel hub information database to automatically retrieve the precise parameters of the wheel hub, which are then sent to the flexible clamping mechanism and intelligent tightening system for pre-configuration. The infrared depth imaging sensor monitors the relative position of the wheel hub and the spindle end face in real time, calculates the alignment deviation, and controls the adjustment mechanism until the deviation meets a set threshold. Then, the flexible clamping mechanism, based on preset database parameters, drives a dual-axis motor to adjust the position of the jaws, automatically adapting to the inner and outer diameters of the wheel hub for initial non-rigid fixing. A servo motor drives the bolts to tighten according to a set torque and cross sequence, providing real-time torque feedback. During tightening, the charge output of each piezoelectric ceramic sheet is collected in real time, converted into a voltage signal by a charge amplifier, and calibrated as a pressure value. The pressure of the same layer of sensors is decomposed using a Fourier series, and the amplitude of the harmonic components related to the number of bolts is extracted as a measure of the clamping torque deviation. An installation error vector is formed based on the alignment error and the clamping torque deviation. A grating encoder integrated into the spindle end face provides time and angle references, and a piezoelectric array is used to collect vibration signals. A time-domain-frequency domain joint solution algorithm based on Hilbert transform is applied to the piezoelectric signals to search for peaks in the envelope signal. For each detected peak moment, the corresponding spindle angle is obtained through encoder interpolation. The average value of multiple cycle detection results is extracted as the valve stem angle position, and the valve stem angle position is defined as... Phase reference.

[0007] Step S2: Synchronously acquire multi-dimensional vibration signals using a configurable multi-sensor array.

[0008] Preferably, based on the wheel hub parameters and installation error vector matched from the database, the optimal sensor combination is automatically selected to achieve dynamic configuration of the sensor array. The sensor array includes: radial vibration, axial vibration, torque, and temperature sensors, and all sensors are configured via an FPGA programmable interface, including range, filtering parameters, and sampling rate. A defined valve stem phase reference is used as the synchronization trigger point; when the encoder detects that the wheel hub has rotated to a certain position... Upon positioning, a hardware trigger signal is issued to initiate synchronous acquisition by all sensors. During acquisition, encoder signals are monitored in real time, and vibration signals are correlated with angle information. The angle value at each sampling point is obtained by encoder interpolation. Based on the encoder's current rotational speed and the hub type (including rigid / flexible) initially determined by database parameters and modal recognition, the sampling rate is dynamically adjusted. The spindle is accelerated according to a preset rotational speed curve, and data is acquired at multiple steady-state rotational speed points, including low-speed, high-speed, and points near the critical rotational speed. During acquisition, the signal-to-noise ratio, peak factor, and kurtosis index of each channel signal are calculated in real time. When the signal quality of a certain channel is substandard, the gain of that sensor is automatically adjusted or re-acquisition is triggered. Using the obtained installation error vector, the impact of installation error on the vibration signal is calculated in real time through a pre-calibrated transfer function, and preliminary compensation is performed during acquisition. All acquired raw data undergoes preliminary preprocessing, and the packaged data is transmitted to the host computer via a high-speed PCIe interface, while a backup is stored locally.

[0009] Step S3: Adaptively decompose the original mixed signal into physically meaningful modal components and eliminate environmental vibration interference.

[0010] Preferably, the vibration signals, including acceleration and displacement, acquired synchronously from multiple channels, along with angle information calibrated using a step phase reference, are used as signal inputs. Initial parameters for the VMD algorithm, including the number of modes, penalty factor, and noise tolerance, are set adaptively based on the characteristics of the acquired signals and the known parameters of the wheel hub. The number of modes is determined using an adaptive method. The goal of synchronously setting VMD is to decompose the original signal into modal functions, minimizing the sum of the estimated bandwidths of each mode, and ensuring that the sum of all modes equals the original signal. After decomposition, modal components are obtained, each corresponding to a different frequency range. Environmental vibration frequencies, installation error frequencies, and unbalanced excitation frequencies are used as prior information for filtering. The filtered modal components are then superimposed to obtain a preliminarily purified signal.

[0011] Step S4: Use the FastICA-RLS hybrid algorithm to separate independent source signals that are strongly correlated with the actual imbalance from the pre-purified signal, and track and compensate for non-stationary interference caused by hub shape and position deviation in real time.

[0012] Preferably, by superimposing all unbalanced correlated mode components of each channel, a composite unbalanced signal and an observed signal for each channel are obtained. The FastICA-RLS hybrid algorithm is then used to separate the independent source signal containing the unbalanced source signal and the interference signal from the observed signal.

[0013] Preferably, the separation of independent source signals using the FastICA-RLS hybrid algorithm includes: blind source separation using the FastICA hybrid algorithm. From the separated independent components, the unbalanced source signal and the form-position deviation interference signal are identified. The spectrum of each source signal is calculated, and correlation analysis is performed with the frequency conversion. Simultaneously, the cross-correlation with the installation error reference signal is calculated. Form-position deviation interference is tracked and eliminated in real time from the observed signal. The form-position deviation interference source signal is set as the input signal, and the interference component in any channel of the observed signal or the mixed signal after FastICA separation is set as the desired signal. The form-position deviation interference source signal is used as the input to the RLS filter, and the observed signal is used as the desired signal to estimate the propagation path of the interference in the observed signal. The interference component is subtracted from the observed signal to obtain the clean unbalanced signal. Finally, the clean unbalanced signal, the estimated form-position deviation interference, the calculated environmental vibration suppression rate, and the form-position deviation measurement error are output.

[0014] Step S5: Construct a flexible rotor model based on modal parameter identification and calculate the unbalance.

[0015] Preferably, the vibration response data obtained through multi-point synchronous measurement using a piezoelectric ceramic array, after step purification, is used as input. A characteristic system implementation algorithm is applied to construct a Hankel matrix. Singular value decomposition is performed on the Hankel matrix, and the system order is determined based on the magnitude of the singular values. Observable and controllable matrices are extracted. A discrete state space matrix is ​​constructed, and the eigenvalues ​​and eigenvectors of the space matrix are calculated to obtain the system's natural frequencies, damping ratios, and mode shapes. When modal parameters need to be re-identified, the ERA algorithm is used for modal parameter re-identification based on the pure unbalance signal. Rigid and flexible rotors are determined based on the obtained critical speed and the current operating speed. For rigid rotors, counterweights are applied on two correction planes, and a balance equation is constructed. For flexible rotors, an extended influence coefficient matrix is ​​constructed. A preliminary influence coefficient matrix is ​​calculated using a finite element model and corrected using measured data. Finally, the obtained unbalance is decomposed into unbalance mass and phase on two planes for output.

[0016] Step S6: Based on multi-source information fusion, adaptive Kalman filter error compensation and prediction are performed, and the compensated imbalance is output.

[0017] Preferably, a state vector is defined, a historical database is constructed to store historical data for each detection, and an adaptive Kalman filter is used to adjust the noise covariance matrix based on real-time data. The filtered output is a component in the state estimate, which is the compensated imbalance. Simultaneously, the error term in the state estimate is used to compensate for the current measurement, and the state equation is used to predict the future trend of imbalance and the development trend of various errors.

[0018] Step S7: The optimal counterweight scheme is automatically calculated based on the compensated imbalance and displayed and guided intuitively through the human-machine interface.

[0019] Preferably, a load-free identification algorithm is used to identify the compensated imbalance output. Combined with the wheel hub's CAD model and material database, the software simulates the effects of adding / removing counterweights at different locations. Under the constraints of allowable counterweight positions and standard counterweight mass sequences, the objective function is constructed to minimize the residual unbalanced force / couple, with the constraints being the minimum number of counterweights and accessible installation positions. Integer programming or heuristic algorithms are used for rapid solution to obtain a specific counterweight scheme. Based on the calculated residual unbalance, automatic judgment is made against the target accuracy standard, and the result is displayed on the industrial control computer screen. Finally, the original detection data, processing procedures, final results, and counterweight scheme are stored in the database to provide historical reference for subsequent inspections of the same model of wheel hub.

[0020] Compared with the prior art, the beneficial effects achieved by the present invention are: (1) This invention integrates the establishment of installation reference and signal acquisition on the spindle to realize detection and calibration. It combines information from multiple sensors such as vibration, acoustics, and temperature to perform adaptive synchronous acquisition. At the same time, it introduces physical constraints in the signal decomposition, separation, and modeling steps to improve the physical interpretability of the processing.

[0021] (2) This invention achieves deep separation of real unbalanced signals by combining the FastICA-RLS hybrid algorithm with physical constraints, and combines physical models with neural networks, achieving both physical accuracy and data adaptability. Real-time, batch, and long-term multi-level compensation is achieved through multi-scale Kalman filtering compensation, and the reliability of the results is improved by combining online learning and historical data fusion.

[0022] (3) This invention significantly reduces the impact of interference such as installation errors and environmental vibrations through multi-source information fusion and intelligent algorithms, thereby improving detection accuracy. At the same time, through automated installation, calibration and detection, manual intervention is reduced and detection efficiency is improved. It can also adapt to different wheel hub models and different working conditions, enhancing the system's adaptability. Finally, through adaptive compensation and online learning, it achieves long-term stability and robustness. Attached Figure Description

[0023] Figure 1 This is a flowchart of the steps of an automated detection method for dynamic balance parameters of automobile wheel hubs provided in an embodiment of the present invention. Detailed Implementation

[0024] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention are within the scope of protection of the present invention.

[0025] Example 1: like Figure 1 As shown in the figure, the automated detection method for dynamic balance parameters of automobile wheel hubs described in this embodiment includes the following steps: Step S1: Using piezoelectric ceramic array self-sensing spindle technology, the detection benchmark establishment and signal acquisition are integrated into the spindle body, and the hub installation process is merged with the detection benchmark establishment process.

[0026] In this embodiment, the present invention integrates the physical installation process of the wheel hub with the establishment process of the testing benchmark, transferring the benchmark establishment function from external tooling to the main shaft body, realizing testing as calibration. It uses the piezoelectric effect to replace the traditional mechanical alignment mechanism, eliminating mechanical wear and clearance errors, breaking through the traditional paradigm of installation before testing. Through the self-sensing of the main shaft, it realizes a real-time closed loop between the installation and testing processes, eliminating the influence of benchmark errors on dynamic balance testing from the source, providing accurate phase and spatial benchmarks for subsequent testing, and improving testing accuracy.

[0027] For example, the specific implementation process includes: Pre-configuration and database matching: In this invention, the appearance of the wheel hub to be tested is scanned by an infrared depth imaging sensor to obtain feature images such as bolt holes and center holes. The image features are matched with the wheel hub information database to automatically retrieve the precise parameters of the wheel hub (bolt hole pitch circle diameter, eccentricity, center hole diameter, etc.) and send them to the flexible clamping mechanism and intelligent tightening system for pre-configuration.

[0028] Visual-guided alignment and flexible automatic clamping: An infrared depth imaging sensor monitors the relative position of the hub and the spindle end face in real time, calculates the alignment deviation, and controls the adjustment mechanism until the deviation meets the requirements (alignment error). Marking error Then, the flexible clamping mechanism drives the dual-axis motor to adjust the position of the jaws according to the preset database parameters, automatically adapting to the inner and outer diameters of the wheel hub to complete the initial non-rigid fixation.

[0029] Intelligent tightening and baseline establishment: Servo tightening: In this invention, the bolts are driven by a servo motor to tighten according to the set torque (based on the database or vehicle standard) and cross sequence, and the torque value is fed back in real time to ensure torque uniformity (difference within ±2%).

[0030] Piezoelectric Array Self-Sensing: During the process of mounting the hub onto the spindle tapered surface and tightening it, factors such as the fit error between the hub's center hole and the spindle tapered surface, bolt hole alignment deviations, and uneven tightening torque can lead to uneven contact pressure distribution between the hub and the spindle. This uneven pressure distribution can cause slight tilting and eccentricity of the hub, i.e., centering error (including radial displacement and tilt angle) and clamping torque deviation (uneven preload of each bolt). This invention embeds 24 PZT-5H piezoelectric ceramic sheets on the spindle tapered surface in three layers (axially distributed) and eight directions (uniformly distributed circumferentially), thus each layer has 8 sensors, located at... Orientation. During the tightening process, the charge output of each piezoelectric ceramic sheet is collected in real time, converted into a voltage signal by a charge amplifier, and calibrated as a pressure value. Let the first... layer , No. position The pressure measured by the sensor is Ideally, when the hub is perfectly aligned and the torque is uniform, the pressure measured by each sensor should be equal (or symmetrically distributed). The non-uniformity of the actual pressure distribution reflects installation errors. Therefore, the contact between the hub and the spindle is simplified to an elastic contact problem. According to Hertzian contact theory, under the assumption of small displacements, the pressure distribution and displacement can be approximated as a linear relationship. Let the installation error parameters be: Radial eccentricity: In the coordinate system of the spindle end face, the radial displacement of the hub center relative to the spindle center is... and (Unit: mm)

[0031] Tilt angle: The angle between the hub axis and the spindle axis. plane and The projection of the plane is and (Unit: radians)

[0032] Clamping torque deviation: Uneven preload of each bolt leads to uneven pressure distribution on the hub end face, which can be equivalent to... The shaft torque is uneven, but it mainly affects the circumferential pressure distribution.

[0033] The pressure distribution is a function of these error parameters. Since the sensors are discretely distributed, this invention uses linear interpolation or fitting to establish the relationship. Let the circumferential angle at each sensor location be... axial position is ( Starting from the end face of the spindle and moving inwards, the following are: When perfectly aligned and without tilt, the theoretical pressure at each sensor position is... (Uniform distribution), by initializing theoretical pressure The average pressure from all sensors is used to calculate the theoretical pressure change at each sensor location, yielding the alignment error parameter: Radial displacement: Direction angle: Inclination angle: , tilt direction, etc.

[0034] Clamping torque deviation is mainly reflected in the non-uniformity of circumferential pressure distribution, especially the pressure difference among the eight sensors in the same layer. In this invention, clamping torque deviation is defined as the harmonic component of pressure distribution caused by the uneven preload of each bolt. Simultaneously, due to the different bolt preloads, the pressure distribution will exhibit periodic changes, the period of which is related to the number of bolts (for example, for a 5-bolt hub, a 5th harmonic will appear). In this invention, the pressure of the eight sensors in the same layer is decomposed into a Fourier series, and the amplitude of the harmonic component related to the number of bolts is extracted as a measure of clamping torque deviation. Finally, an installation error vector is formed based on the centering error and clamping torque deviation.

[0035] Phase self-calibration: Simultaneously, the spindle begins to rotate at a low speed (e.g., 30-60 RPM), providing accurate time and angle references through a grating encoder (36,000 pulses per revolution) integrated on the spindle end face, and using a piezoelectric array to collect vibration signals.

[0036] Phase reference identification algorithm: A time-domain-frequency domain joint solution algorithm based on Hilbert transform is applied to the piezoelectric signal. Since the valve stem is a significant local mass point, it induces a specific phase response when passing through the sensor. The algorithm searches for peaks in the envelope signal. Because the valve stem generates a significant impact with each rotation, there is a main peak in the envelope within each rotation cycle. For each detected peak moment, the corresponding spindle angle is obtained through encoder interpolation. Since the valve stem is a fixed point on the hub, its angular position relative to the spindle should be fixed (ignoring spindle torsional deformation, etc.). Therefore, the angle values ​​detected in multiple cycles should be basically consistent. In this invention, the average value of the detection results from multiple cycles is extracted as the valve stem angle position. The valve stem angle position is defined as... The phase reference, i.e., the phase of the subsequently detected imbalance, is relative to the valve stem angle. For example, if the phase of the imbalance is... This indicates that the unbalanced mass point is located at the point where the valve stem rotates counterclockwise. The angle position does not require manual marking.

[0037] Step S2: Synchronously acquire multi-dimensional vibration signals using a configurable multi-sensor array.

[0038] In this embodiment, based on the established accurate benchmark and error compensation, the present invention uses a configurable multi-sensor array to synchronously collect multi-dimensional dynamic response signals under the state of wheel hub rotation, providing high-quality and highly synchronous raw data for subsequent signal processing.

[0039] For example, the specific implementation process includes: Dynamic sensor array configuration: The optimal sensor combination is automatically selected based on wheel hub parameters matched from the database and the installation error vector. For example, if the installation error indicates significant radial eccentricity, the weight of the radial acceleration sensor is increased, and its range is adjusted. The sensor array includes: Radial vibration: 3 MEMS accelerometers (range selectable ±50g, ±100g, ±200g, dynamically selected based on the unbalanced force estimated from the hub mass); Axial vibration: 2 eddy current displacement sensors (range 0-2mm, resolution) ); Torque sensor: One non-contact torque sensor for monitoring spindle drive torque fluctuations, with a sampling rate ≥5kHz; Temperature sensors: Four PT100 temperature sensors are installed near the spindle bearing, drive motor, environment, and wheel hub, respectively, for temperature drift compensation; All sensors are configured via an FPGA programmable interface, including range, filtering parameters, sampling rate, etc.

[0040] Synchronous triggering and acquisition: based on a defined valve stem phase reference ( (Position) serves as the synchronous trigger point; when the encoder detects that the wheel hub has rotated to... Upon positioning, a hardware trigger signal is issued to initiate synchronous acquisition by all sensors. This invention employs an FPGA to achieve hardware-level synchronization, ensuring that the sampling clocks of all channels are from the same source, with a synchronization error of less than 10ns. Furthermore, the encoder signal is monitored in real time during acquisition, mapping the vibration signal to angle information; the angle value at each sampling point is obtained through encoder interpolation.

[0041] Adaptive sampling rate adjustment: based on the encoder's current rotational speed. Based on the initial judgment of wheel hub type (rigid / flexible) using database parameters and modal identification, the sampling rate is dynamically adjusted, and oversampling technology is employed. The base sampling rate... Set to:

[0042] in, This represents the harmonic order (usually taken as 10). This is the highest analysis frequency corresponding to the current rotational speed (and includes at least the first three critical rotational speeds). In the high-speed phase ( A higher sampling rate (e.g., 20 kHz) is used to capture the higher-order modes of the flexible rotor; a lower sampling rate (e.g., 5 kHz) is used in the low-speed phase to reduce the amount of data.

[0043] Steady-state data acquisition at multiple speed points: In this invention, the spindle is controlled to accelerate according to a preset speed curve, and data is acquired at multiple steady-state speed points, including: Low speed point (e.g., 300 RPM): used for rigid rotor balance analysis; High-speed point (e.g., operating speed, 2000-3000 RPM): used for flexible rotor balance analysis; Points near the critical speed (if the operating speed is close to the critical speed): used for accurate identification of modal parameters.

[0044] At each steady-state rotational speed point, at least 10 complete rotational cycles are collected to ensure the statistical reliability of subsequent signal processing.

[0045] Real-time signal quality monitoring and feedback: During the acquisition process, the signal-to-noise ratio (SNR), peak factor, kurtosis, and other indicators of each channel signal are calculated in real time. If the signal quality of a certain channel is substandard (e.g., SNR < 20dB), the system automatically adjusts the gain of that sensor or triggers a re-acquisition. Using the obtained installation error vector, the impact of installation error on the vibration signal is calculated in real time through a pre-calibrated transfer function, and preliminary compensation is performed during the acquisition process to reduce the burden of subsequent processing.

[0046] Data preprocessing and packaging: The acquired raw data undergoes preliminary preprocessing, including: removing DC components, applying an anti-aliasing filter (implemented in hardware), and linearizing the temperature sensor data. The multi-channel data is then packaged according to a unified timestamp and angle stamp. Each data packet contains: timestamp, angle value, data from each sensor channel, encoder raw pulse count, temperature data, and installation error compensation flag.

[0047] Data transmission and storage: The packaged data is transmitted to the host computer (DSP part) via a high-speed PCIe interface for further processing, and backup storage is performed locally. The storage format adopts the standard HDF5 format, which facilitates subsequent offline analysis and data mining.

[0048] Step S3: Adaptively decompose the complex original mixed signal into a series of physically meaningful modal components, and initially eliminate obvious environmental vibration interference.

[0049] In this embodiment, the specific implementation process includes: Signal Input and Initialization: Vibration signals, including acceleration and displacement, acquired synchronously from multiple channels, along with angle information calibrated using a step phase reference, are used as signal inputs. Simultaneously, the initial parameters of the VMD algorithm, including the number of modes, are set. Punishment factor Noise tolerance The system is adaptively initialized based on the characteristics of the acquired signal (such as spectral features) and the known parameters of the wheel hub (such as wheel hub size and mass).

[0050] Adaptive determination of the number of modes Due to the varying signal complexity at different wheel hubs and rotational speeds, the fixed number of modes... The value may lead to over-decomposition or under-decomposition. This invention uses an adaptive method to determine the number of modes. : a. Calculate the power spectral density (PSD) of the input signal and identify significant peaks (peak heights exceeding three standard deviations of the average PSD). b. Based on the number of significant peaks, the natural frequency of the hub, and the rotational speed information, a preliminary estimate of the number of modes is made. Value range (usually 5-8).

[0051] c. Using the center frequency observation method: from Initially, the number of modes is increased sequentially. This continues until modes with similar center frequencies (frequency difference less than 5 Hz) appear or meaningless modes appear (energy percentage less than 1%), at which point the number of modes is... That is the optimal value.

[0052] Variational Mode Decomposition (VMD) is specifically implemented by decomposing the original signal into mode functions. This minimizes the sum of the estimated bandwidths for each mode, and the sum of the bandwidths for all modes equals the original signal.

[0053] Modal component selection and reconstruction: obtaining modal components after decomposition Each component corresponds to a different frequency range, and the environmental vibration frequency, installation error frequency, and unbalanced excitation frequency are used as prior information for filtering. a. Environmental vibration components: These typically have low frequencies (0-5Hz) and are independent of rotational speed. They can be identified by the frequency range of the environmental vibration components. b. Installation error component: The frequency is synchronized with the rotational speed (1st harmonic) and is related to the calculated installation error. It can be identified by calculating the 1st harmonic amplitude of each component and performing correlation analysis with the installation error vector.

[0054] c. Unbalanced components: These typically include first harmonics, second harmonics, etc., and are related to rotational speed, with relatively high energy.

[0055] d. Other interferences: such as resonance, which occurs at higher frequencies.

[0056] Screening criteria: 1. Calculate the center frequency of each modal component. ,if If the vibration frequency is within the range of 0-5Hz and is independent of the rotational speed, it will be eliminated.

[0057] 2. Calculate the first harmonic amplitude of each modal component. ,if If the magnitude of the installation error is proportional to the calculated installation error (radial eccentricity in the installation error vector) (the proportionality coefficient is obtained through calibration), then this component is marked as the installation error component, and its amplitude and phase are recorded for subsequent compensation.

[0058] 3. Among the remaining components, the components with a center frequency of 20-100Hz (the main frequency band of unbalanced excitation) and high energy are selected as candidate unbalanced components for subsequent analysis.

[0059] Reconstructed signal: The filtered modal components (mainly candidate unbalanced components) are superimposed to obtain the preliminarily purified signal. .

[0060] For example, in this invention, when the acquired radial acceleration signal is The sampling frequency was 10kHz, and the rotation speed was 1200rpm (20Hz). PSD calculations revealed significant peaks at harmonics of 20Hz, 40Hz, and 60Hz, with peaks also observed below 5Hz. A preliminary estimate of the number of modes was made. (Including 1 environmental vibration mode, 1 installation error mode, 3 unbalanced harmonic modes, and 1 noise mode). VMD decomposition was performed, yielding 6 modal components. The mode with a center frequency of 5Hz (environmental vibration) was removed; the mode with a center frequency of 20Hz was correlated with the calculated installation error (the correlation coefficient between its first harmonic amplitude and the radial eccentricity of the installation error vector was >0.8), and was marked as an installation error component, with its amplitude and phase recorded; the modes with center frequencies of 20Hz, 40Hz, and 60Hz had higher energy and were correlated with the rotational frequency harmonics, so they were retained as candidate unbalanced components. Finally, the modal components of 20Hz, 40Hz, and 60Hz were superimposed to obtain the preliminarily purified signal. .

[0061] Step S4: Use the FastICA-RLS hybrid algorithm to separate independent source signals that are strongly correlated with the actual imbalance from the pre-purified signal, and track and compensate for non-stationary interference caused by hub shape and position deviations (such as slight deformation or out-of-roundness) in real time.

[0062] In this embodiment, the present invention obtains a composite unbalanced signal for each channel by superimposing all unbalanced correlated mode components of each channel, and thus obtains the observation signal for each channel. Let the observed signal matrix be , dimension , This represents the number of sampling points.

[0063] For example, from the observed signal Separate statistically independent source signals from The signal contains unbalanced source signals and interference signals. The specific implementation process of FastICA blind source separation includes: a) Centering: Remove the mean to make the signal have zero mean; b) Whitening: The observed signal is whitened using PCA to obtain... ,in It is a whitening matrix; c) Select the number of independent components to be separated (here set to M, the same as the number of channels); d) Initialize the weight vector for each independent component. (Random unit vector); e) Iterative update (using a fixed-point algorithm that maximizes negative entropy): For each ,renew ; f) Orthogonalization: After each iteration, the weight set is... Orthogonalization is performed to prevent convergence to the same component; g) Repeat the iterations until convergence (adjacent iterations) (Change less than the threshold) h) The separation matrix is ​​obtained, and the source signal estimate is obtained. .

[0064] Source signal identification: From the isolated independent components, identify unbalanced source signals that are highly correlated with the rotational speed frequency (or harmonics) and whose energy is mainly concentrated at the rotational speed and its harmonics, as well as form and position deviation interference signals that have non-stationary characteristics and are highly correlated with the installation error reference signal. This is achieved by calculating each source signal... The spectrum is analyzed for correlation with the frequency conversion, and the cross-correlation with the installation error reference signal is calculated.

[0065] RLS adaptive filtering tracks form and position deviation interference: This invention tracks and eliminates form and position deviation interference from the observed signal in real time. The first identified source signal is defined as the unbalanced source signal, the second as the form and position deviation interference signal, and the rest as noise. The form and position deviation interference source signal (i.e., the second source signal) is set as the input signal (reference signal), and the interference component in any channel of the observed signal or the mixed signal after FastICA separation is set as the desired signal. The form and position deviation interference source signal is used as the input of RLS, and the observed signal is used as the desired signal to estimate the propagation path of the interference in the observed signal. Then, the interference component is subtracted from the observed signal to obtain the clean unbalanced signal. Specifically, this includes: initializing the RLS filter; updating the input vector, prior error, gain vector, weights, and inverse correlation matrix at each time step; and outputting the interference estimate and the clean signal. In this embodiment, only one channel is RLS filtered, but in practice, the same operation can be performed on each channel, or a multi-channel RLS can be used. That is, this invention uses block processing, processing one data block (e.g., 2048 points) at a time.

[0066] Finally, the output includes the clean unbalanced signal obtained after RLS filtering for each channel, the estimated form and position deviation interference, the calculated environmental vibration suppression rate (the energy ratio of the original observed signal to the clean signal in the environmental vibration frequency band), and the form and position deviation measurement error (compared with the known form and position deviation, if any). Environmental vibration suppression rate = (1 - energy of the pure signal in the environmental vibration frequency band / energy of the original signal in the environmental vibration frequency band) × 100%; Geometric deviation measurement error = (estimated geometric deviation amplitude - actual geometric deviation amplitude) / actual geometric deviation amplitude × 100%.

[0067] Step S5: Construct a flexible rotor model based on modal parameter identification and calculate the unbalance.

[0068] In this embodiment, the present invention further determines whether the hub-spindle rotor is a rigid or flexible rotor at the current rotational speed, and uses a corresponding balance model to accurately calculate the magnitude and phase of the imbalance. Specifically: Modal parameter identification: The vibration response data obtained by multi-point synchronous measurement using a piezoelectric ceramic array and then purified is used as input. The Feature Implementation Algorithm (ERA) is applied to construct the Hankel matrix. and ,right Perform singular value decomposition (SVD) to determine the system order based on the magnitude of the singular values. Furthermore, observable and controllable matrices are extracted, and discrete state space matrices are constructed. , , ,calculate The eigenvalues ​​and eigenvectors yield the system's natural frequencies, damping ratios, and mode shapes (usually the first three modes are extracted). When accurate modal parameters are obtained through the ERA algorithm, they are used directly. When re-identification is required, the ERA algorithm is used to re-identify the modal parameters based on the pure unbalanced signal.

[0069] Balance model selection and construction: based on the obtained critical speed (Rotation speed corresponding to the first natural frequency) and current operating speed Determining whether a rotor is rigid or flexible: when When the rotor is determined to be rigid, the rigid rotor balance model (two-sided influence coefficient method) is adopted. when When the rotor is determined to be flexible, the flexible rotor balance model (modal influence coefficient method) is adopted.

[0070] Constructing the equilibrium equations: For a rigid rotor, counterweights are applied to two correction planes (usually the two sides of the hub), and the equilibrium equations are as follows: Where [C] is a 2×2 influence coefficient matrix (vibration response caused by each plane at a measuring point). These are the unbalanced quantities (complex vectors, including magnitude and phase) on the two planes to be determined. This is the vibration response (complex vector) measured at two measuring points (usually two radial sensors). For a flexible rotor, an extended influence coefficient matrix is ​​constructed. This invention defines measurements at P rotational speeds, with two measuring points at each speed. Therefore, the response vector is 2P-dimensional. The unbalance is still an unbalance on two planes (assuming balance only on two planes). Thus, the influence coefficient matrix is ​​2P×2-dimensional, and the balance equation is: ,in It is a matrix of influence coefficients that vary with rotational speed. It is the vibration response at different rotational speeds.

[0071] Obtaining the influence coefficient matrix: a) Establish a finite element model of the hub-spindle system and update the model using the obtained modal parameters (natural frequency, damping ratio, mode shape); b) Apply a unit unbalance (e.g., 1 g·mm) to the two correction planes of the model and calculate the vibration response (amplitude and phase) at the two measuring points at different rotational speeds. c) Obtain the influence coefficient matrix .

[0072] In practical applications, this invention employs a hybrid approach: a preliminary influence coefficient matrix is ​​calculated using a finite element model, and then corrected using measured data (pure unbalanced response).

[0073] Unbalance solution: For a rigid rotor, the balance equations are a linear system of equations, which can be solved directly: For flexible rotors, since the number of equations exceeds the number of unknowns (2P>2), this invention uses the least squares method to solve them: ,in This indicates the conjugate transpose.

[0074] Output the imbalance: Finally, output the imbalance obtained from the solution. Decompose the output into unbalanced mass and phase on two planes: Plane 1: Mass m1, Phase (Relative to valve stem phase reference); Plane 2: Mass m2, Phase .

[0075] Step S6: Based on multi-source information fusion, adaptive Kalman filter error compensation and prediction are performed, and the compensated imbalance is output.

[0076] In this embodiment, the present invention performs real-time prediction and compensation for systematic errors introduced by installation errors, sensor temperature drift, long-term operational drift, etc., thereby improving the repeatability and long-term stability of detection. Specifically, it includes: State-space model enhancement: Defines the state vector including actual imbalance, installation error vector, slowly varying drift errors such as temperature drift, model error, and model error.

[0077] Multi-source historical data fusion: This invention constructs a historical database storing the following information from each detection: installation error estimation, phase reference, raw signals (or eigenvalues) from multiple sensors, temperature data, IMF components after VMD decomposition, modal parameters, separated true imbalance signals, interference signals, model parameters, imbalance estimates, uncertainty information, and temperature and humidity, etc. Utilizing this historical data: a. Initialize the state estimate and covariance matrix: Initialize using the statistical properties of historical data; b. Learning the state transition model: Through time series analysis, learn the changing patterns of state variables, thereby adjusting the state transition matrix F and the process noise covariance matrix Q.

[0078] c. Learn the characteristics of observation noise: Adjust the observation noise covariance matrix R based on the difference between the observed values ​​and the state estimates in the historical data.

[0079] Adaptive Kalman Filtering Algorithm: This invention employs adaptive Kalman filtering, adjusting the noise covariance matrix based on real-time data.

[0080] Real-time compensation and prediction: The components in the filtered output state estimate are the compensated imbalance. Simultaneously, this invention utilizes error terms in the state estimate (such as installation error and sensor drift) to compensate for the current measurement. Furthermore, it predicts the future state through the state equation, i.e., the trend of imbalance change and the development trend of various errors over a future period.

[0081] Performance Evaluation: The performance of this invention is evaluated using the following indicators. Convergence rate of state estimation: the time response from start to estimated stability; Estimation accuracy: the deviation from the actual imbalance (when the actual value exists); Repeatability error: the coefficient of variation of multiple tests conducted under the same conditions; Prediction accuracy: the error in predicting future changes in imbalance.

[0082] Step S7: The optimal counterweight scheme is automatically calculated based on the compensated imbalance and displayed and guided intuitively through the human-machine interface.

[0083] In this embodiment, the compensated imbalance is identified using a load-free identification algorithm, combined with the CAD model of the wheel hub and a material database, to simulate the effect of adding / removing counterweights at different locations in the software. An optimization problem is constructed under the constraints of allowed counterweight locations (e.g., a specific area inside the wheel flange) and a standard counterweight mass sequence. The objective function in this invention is to minimize the residual unbalanced force / couple, with constraints including minimizing the number of counterweights and ensuring the installation locations are accessible. Integer programming or heuristic algorithms (such as genetic algorithms) are then used for rapid solution to obtain a specific counterweight scheme (the mass and angle of the counterweights to be added to each correction plane). Based on the calculated residual unbalance, an automatic judgment is made against the target accuracy standard, and the following is displayed on the industrial control computer screen: the magnitude and phase of the unbalance (displayed in polar coordinates), the specific counterweight scheme (text and graphic indicators, with the counterweight positions highlighted on the wheel hub image), the predicted post-balancing effect (e.g., amplitude reduction rate, target >85%), and a quality report containing all key parameters and judgment results of this test. Finally, the raw test data, processing procedures, final results, and counterweight scheme are stored in the database to provide historical reference for subsequent tests of the same model of wheel hubs and to trace production quality.

[0084] It should be noted that, in this invention, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus.

[0085] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. An automated detection method for dynamic balance parameters of automobile wheel hubs, characterized in that: The following steps are included: Step S1: Using piezoelectric ceramic array self-sensing spindle technology, the detection benchmark establishment and signal acquisition are integrated into the spindle body, and the hub installation process is merged with the detection benchmark establishment process; Step S2: Synchronously acquire multi-dimensional vibration signals using a configurable multi-sensor array; Step S3: Adaptively decompose the original mixed signal into physically meaningful modal components to eliminate environmental vibration interference; Step S4: Use the FastICA-RLS hybrid algorithm to separate independent source signals that are strongly correlated with the actual imbalance from the initially purified signal, and track and compensate for non-stationary interference caused by hub shape and position deviation in real time. Step S5: Construct a flexible rotor model based on modal parameter identification and calculate the unbalance. Step S6: Based on multi-source information fusion, adaptive Kalman filter error compensation and prediction are performed, and the compensated imbalance is output. Step S7: The optimal counterweight scheme is automatically calculated based on the compensated imbalance and displayed and guided intuitively through the human-machine interface.

2. The automated detection method for dynamic balance parameters of automobile wheel hubs according to claim 1, characterized in that: The implementation of step S1 includes: The infrared depth imaging sensor scans the appearance of the wheel hub under test to obtain feature images of bolt holes and center holes. The image features are matched with the wheel hub information database to automatically retrieve the precise parameters of the wheel hub and send them to the flexible clamping mechanism and intelligent tightening system for pre-configuration. The relative position between the wheel hub and the spindle end face is monitored in real time using an infrared depth imaging sensor. The centering deviation is calculated and the adjustment mechanism is controlled until the deviation meets the set threshold. Then, the flexible clamping mechanism drives the dual-axis motor to adjust the position of the chuck according to the preset database parameters, automatically adapting to the inner and outer diameters of the wheel hub for initial non-rigid fixation. The bolts are tightened by a servo motor according to a set torque and cross sequence, and the torque value is fed back in real time. During the tightening process, the charge output of each piezoelectric ceramic sheet is collected in real time, converted into a voltage signal by a charge amplifier, and calibrated as a pressure value. The pressure of the same layer of sensors is decomposed into Fourier series, and the amplitude of the harmonic components related to the number of bolts is extracted as a measure of the clamping torque deviation. An installation error vector is formed based on the centering error and the clamping torque deviation. A grating encoder integrated into the spindle end face provides time and angle references, and a piezoelectric array is used to collect vibration signals. A time-domain-frequency domain joint solution algorithm based on Hilbert transform is applied to the piezoelectric signals to search for peaks in the envelope signal. For each detected peak moment, the corresponding spindle angle is obtained by encoder interpolation. The average value of multiple cycle detection results is extracted as the valve stem angle position, and the valve stem angle position is defined as... Phase reference.

3. The automated detection method for dynamic balance parameters of automobile wheel hubs according to claim 2, characterized in that: The implementation of step S2 includes: Based on the wheel hub parameters and installation error vector matched from the database, the optimal sensor combination is automatically selected to achieve dynamic configuration of the sensor array. The sensor array includes: radial vibration, axial vibration, torque sensor, and temperature sensor. All sensors are configured through the FPGA programmable interface, including range, filtering parameters, and sampling rate. Using a defined valve stem phase reference as the synchronization trigger start point, when the encoder detects that the wheel hub has rotated to... When the position is determined, a hardware trigger signal is issued to initiate synchronous acquisition of all sensors. During the acquisition process, the encoder signal is monitored in real time, and the vibration signal is correlated with the angle information. The angle value of each sampling point is obtained by encoder interpolation. The sampling rate is dynamically adjusted based on the encoder's current rotational speed and the wheel hub type (including rigid / flexible) initially determined by database parameters and modal recognition. The spindle is controlled to accelerate according to a preset speed curve, and data is collected at multiple steady-state speed points, including: low speed point, high speed point, and point near critical speed. During the acquisition process, the signal-to-noise ratio, peak factor, and kurtosis index of each channel signal are calculated in real time. When the signal quality of a certain channel is substandard, the gain of the sensor is automatically adjusted or a reacquisition is triggered. The installation error vector is used to calculate the impact of the installation error on the vibration signal in real time through the pre-calibrated transfer function, and preliminary compensation is performed during the acquisition process. All the collected raw data are pre-processed and then the packaged data is transmitted to the host computer via a high-speed PCIe interface, while a backup is stored on the local machine.

4. The automated detection method for dynamic balance parameters of automobile wheel hubs according to claim 3, characterized in that: The implementation of step S3 includes: Vibration signals, including acceleration and displacement, acquired synchronously from multiple channels, as well as angle information calibrated using a step phase reference, are used as signal inputs. The initial parameters of the VMD algorithm are set, including the number of modes, penalty factor, and noise tolerance. The initialization is adaptively performed based on the characteristics of the acquired signal and the known parameters of the hub. The number of modes is determined using an adaptive method. The goal of synchronously setting VMD is to decompose the original signal into mode functions, minimizing the sum of the estimated bandwidths of each mode, and ensuring that the sum of all modes equals the original signal. After decomposition, modal components are obtained, each corresponding to a different frequency range. The environmental vibration frequency, installation error frequency, and unbalanced excitation frequency are used as prior information for screening. The filtered modal components are superimposed to obtain the preliminarily purified signal.

5. The automated detection method for dynamic balance parameters of automobile wheel hubs according to claim 4, characterized in that: Determining the number of modes using an adaptive method includes: Calculate the power spectral density of the input signal and identify significant peaks; Based on the number of significant peaks, the natural frequency of the hub, and the rotational speed information, the range of the number of modes is initially estimated; Using the center frequency observation method, the number of modes is increased sequentially until a mode with a similar center frequency or a predefined meaningless mode appears. The number of modes at this point is the optimal value, and it is then output.

6. The automated detection method for dynamic balance parameters of automobile wheel hubs according to claim 5, characterized in that: In step S4, by superimposing all unbalanced correlated mode components of each channel, a composite unbalanced signal and an observed signal for each channel are obtained. The FastICA-RLS hybrid algorithm is then used to separate the independent source signal containing the unbalanced source signal and the interference signal from the observed signal.

7. The automated detection method for dynamic balance parameters of automobile wheel hubs according to claim 6, characterized in that: The FastICA-RLS hybrid algorithm is used to separate independent source signals, including: Blind source separation is performed using a hybrid FastICA algorithm; From the isolated independent components, the unbalanced source signal and the form and position deviation interference signal are identified. By calculating the spectrum of each source signal, correlation analysis is performed with the frequency conversion, and the cross-correlation with the installation error reference signal is also calculated. Real-time tracking and elimination of form and position deviation interference from the observed signal; setting the form and position deviation interference source signal as the input signal; setting any channel of the observed signal or the interference part in the mixed signal after FastICA separation as the desired signal. The form and position deviation interference source signal is used as the input of the RLS filter, the observed signal is used as the desired signal to estimate the propagation path of the interference in the observed signal, and the interference component is subtracted from the observed signal to obtain the pure unbalanced signal. The final output includes a clean unbalanced signal, a form and position deviation interference estimate, an environmental vibration suppression rate, and a form and position deviation measurement error.

8. The automated detection method for dynamic balance parameters of automobile wheel hubs according to claim 7, characterized in that: The implementation of step S5 includes: The vibration response data obtained by multi-point synchronous measurement using a piezoelectric ceramic array after step purification is used as input. The Hankel matrix is ​​constructed by applying the feature system implementation algorithm. The Hankel matrix is ​​decomposed into singular values. The system order is determined according to the magnitude of the singular values, and the observable and controllable matrices are extracted. The discrete state space matrix is ​​constructed, and the eigenvalues ​​and eigenvectors of the space matrix are calculated to obtain the natural frequency, damping ratio and mode shape of the system. When modal parameters need to be re-identified, the ERA algorithm is used to re-identify modal parameters based on the pure unbalanced signal; The determination of whether a rotor is rigid or flexible is made based on the obtained critical speed and the current operating speed. For a rigid rotor, counterweights are applied on two correction planes to construct the equilibrium equations; For flexible rotors, construct an extended influence coefficient matrix; A preliminary influence coefficient matrix was calculated using a finite element model, and then corrected using measured data. Finally, the unbalanced quantity obtained from the solution is decomposed into unbalanced mass and phase on two planes and output.

9. The automated detection method for dynamic balance parameters of automobile wheel hubs according to claim 8, characterized in that: The implementation of step S6 includes: Define the state vector; Build a historical database to store historical data for each detection; An adaptive Kalman filter is used to adjust the noise covariance matrix based on real-time data; The component in the filtered output state estimate is the compensated imbalance quantity. Simultaneously, the error term in the state estimation is used to compensate for the current measurement, and the trend of the unbalance quantity in the future time is predicted through the state equation, as well as the development trend of various errors.

10. The automated detection method for dynamic balance parameters of automobile wheel hubs according to claim 9, characterized in that: Step S7 includes: The unbalanced output is identified using a load-free identification algorithm without trial weights. Combined with the CAD model of the wheel hub and the material database, the effect of adding / removing counterweights at different locations is simulated in the software. Under the constraints of allowable counterweight locations and standard counterweight mass sequence, the objective function is to minimize the residual unbalanced force / couple, with the constraints being the minimum number of counterweights and the accessible installation locations. The specific weighting scheme can be obtained by quickly solving the problem using integer programming or heuristic algorithms. Based on the calculated residual imbalance, the system automatically determines the target accuracy standard and displays the result on the industrial control computer screen. Finally, the raw test data, processing procedures, final results, and counterweight scheme are stored in the database to provide historical reference for subsequent tests of the same model of wheel hub.

Images