Intelligent Weighing Sensor and Method Based on Dual-Plane Beam and High-Frequency Non-Contact Vibration Measurement
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- CHENGDU RING EXPRESSWAY (WESTERN SECTION) CO LTD
- Filing Date
- 2026-01-09
- Publication Date
- 2026-05-26
Smart Images
Figure CN122084073A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of weighing sensor technology, and particularly relates to an intelligent weighing sensor and method based on dual-plane beam and high-frequency non-contact vibration measurement. Background Technology
[0002] Dynamic weighing technology is a crucial component in highway overload control, weight-based toll collection, and logistics management. Its measurement accuracy and reliability directly impact law enforcement fairness and operational efficiency. Traditional weighing sensors are primarily based on the principle of resistance strain, calculating mass by measuring the strain of an elastic body under load. While these sensors are stable in static or quasi-static scenarios, they face multiple technical bottlenecks in real-world dynamic weighing environments. First, the complex mechanical vibrations generated when vehicles pass over the weighing platform at high speeds are superimposed on the effective strain signal, creating significant interference. Due to the limited response frequency of traditional sensors (typically below 1kHz) and the lack of effective vibration separation mechanisms, dynamic measurement errors generally exceed ±1%FS, failing to meet the ever-increasing requirements for measurement accuracy. Second, their functionality is limited; existing sensors can only provide mass or axle load data and cannot identify vehicle axle types (such as drive axles and non-drive axles). External equipment such as infrared light curtains, laser scanning, or image recognition must be used for axle type identification, which not only significantly increases system complexity and deployment costs but also introduces new error sources due to multi-device synchronization issues. Furthermore, traditional solutions lack the ability to perceive the health status of the sensors themselves and the weighing platform structure. Sensors are subjected to impact loads, changes in ambient temperature and humidity, and mechanical wear over a long period of time. Their performance may drift slowly or even fail suddenly. Current technology cannot achieve real-time monitoring and early warning of these potential risks. It can only rely on periodic manual calibration and maintenance, which is costly and poses safety hazards.
[0003] Furthermore, in terms of data processing, the single-dimensional signals output by traditional sensors fail to fully extract the rich information contained in vibration signals. For example, changes in structural resonant frequency can reflect early fault characteristics such as bolt loosening and structural fatigue crack propagation. Therefore, developing a new type of weighing sensor that can integrate high-frequency vibration signals, achieve non-contact measurement, and possess multi-parameter intelligent sensing and self-diagnostic capabilities has become an urgent need to overcome the current bottlenecks in dynamic weighing technology and adapt to the development trends of intelligent transportation and the Industrial Internet of Things. Summary of the Invention
[0004] This invention aims to overcome the shortcomings of existing technologies and provide an intelligent weighing sensor and method based on dual-plane beams and high-frequency non-contact vibration measurement. This sensor improves dynamic weighing accuracy through the synchronous acquisition and deep fusion of strain signals and high-frequency vibration signals. It is suitable for applications such as dynamic weighing of vehicles on highways, over-limit detection, and structural health monitoring.
[0005] To achieve the above objectives, the present invention adopts the following technical solution:
[0006] In a first aspect, the present invention provides an intelligent weighing sensor based on dual-plane beam and high-frequency non-contact vibration measurement, the sensor comprising:
[0007] The dual-plane beam, as the force-sensitive core component of the sensor, is made of alloy steel or stainless steel, possessing excellent elastic properties and mechanical strength. The dual-plane beam consists of two parallel beams connected by a connecting structure, capable of withstanding rated loads of 10 to 40 tons.
[0008] The strain sensing unit employs a high-precision resistance strain gauge, which is bonded to the neutral layer or stress concentration area of the biplane beam. The strain gauge has a sensitivity of 1.5 mV / V and a resistance of 350 Ω, and is connected via a Wheatstone bridge, enabling precise measurement of the beam's micro-strain under load.
[0009] The vibration sensing unit employs a nanometer time-grating displacement sensor, which is non-contactly positioned near the elastic deformation region of the biplane beam. Based on the principle of electromagnetic induction, this sensor can acquire the micro-displacement changes of the beam at a sampling frequency greater than 100kHz, achieving a resolution of 0.1μm.
[0010] The temperature sensing unit uses a PT1000 platinum resistance temperature sensor, which is installed at the root of the double-plane beam. The measurement accuracy is ±0.1℃, and it is used to monitor changes in ambient temperature in real time.
[0011] The signal processing module includes a first processing channel, a second processing channel, and a temperature processing unit. The first processing channel processes strain signals and outputs force measurement data; the second processing channel processes vibration signals and outputs vibration characteristic data; both the first and second processing channels are equipped with adaptive gain adjustment circuits, which can dynamically adjust the gain according to the input signal intensity; the temperature processing unit processes temperature data and outputs temperature compensation data.
[0012] The data processing module includes a feature extraction unit, an LSTM neural network model, a health status assessment and alarm unit, and a fault diagnosis unit. The feature extraction unit extracts multi-dimensional feature vectors containing time-domain, frequency-domain, and temporal features from the vibration feature data. The LSTM neural network model is implemented using a three-layer bidirectional LSTM structure, taking force measurement data, multi-dimensional feature vectors, and temperature compensation data as inputs, and outputting mass measurement values, structural health index values, and fault labels. The health status assessment and alarm unit triggers a three-level alarm based on the structural health index value. The fault diagnosis unit identifies the fault type based on the fault label, and fault types include, but are not limited to: strain gauge drift, vibration signal loss, temperature sensor failure, ADC conversion anomaly, power fluctuation, communication interruption, structural fatigue, and bolt loosening.
[0013] The axle type identification module is used to analyze the time-domain waveform of the vibration displacement signal, calculate the vehicle wheelbase based on the peak time interval, extract the spectral energy characteristics of the vibration displacement signal in the 20~50Hz frequency band, and determine the axle type based on the wheelbase and the spectral energy characteristics through a classification model.
[0014] Secondly, the present invention provides an intelligent weighing method based on dual-plane beams and high-frequency non-contact vibration measurement, comprising the following steps:
[0015] Step 1: Synchronously acquire strain signals caused by the load through the strain sensing unit; synchronously acquire vibration signals of the double-plane beam in a non-contact manner through the vibration sensing unit; synchronously acquire ambient temperature data through the temperature sensing unit.
[0016] Step 2: Perform parallel processing on the strain signal, vibration signal, and temperature data to generate force measurement data, vibration characteristic data, and temperature compensation data, respectively;
[0017] Step 3: Extract multidimensional feature vectors from vibration feature data, and input force measurement data, multidimensional feature vectors and temperature compensation data into an LSTM-based neural network model to output mass measurement values, structural health index values and fault labels;
[0018] Step 4: Evaluate the health status based on the structural health index value and trigger the corresponding alarm level;
[0019] Step 5: Analyze the time-domain waveform and frequency-domain characteristics of the vibration signal using the axle type identification module to identify the vehicle axle type.
[0020] Thirdly, the present invention provides a dynamic weighing system that uses the intelligent weighing method and intelligent weighing sensor in the over-limit detection system of a dynamic truck scale on a highway, and can simultaneously output vehicle total weight, axle load, axle type identification information, structural health status information and fault diagnosis information.
[0021] Compared with the prior art, the present invention adopts the following technical solution:
[0022] (1) Through the synergistic effect of real-time acquisition of high-frequency vibration signals and intelligent compensation algorithm, the accuracy bottleneck of traditional dynamic weighing is solved, and the dynamic weighing error is reduced to ≤±0.5%FS, which meets the accuracy requirements of metrology level; the nano-time grating sensor is used to capture microsecond-level vibration transients that traditional sensors cannot respond to, and effectively separates the useful load signal from the vibration interference signal.
[0023] (2) It has the functions of quality measurement, shaft type recognition, structural health monitoring and fault self-diagnosis, which replaces multiple independent devices such as strain sensors, infrared shaft type recognition devices, vibration monitoring instruments and fault diagnosis modules required in traditional systems; it reduces costs and improves system reliability and anti-interference ability.
[0024] (3) The first autonomous axle type identification based on vibration signal was proposed, which got rid of the dependence on infrared beam detectors, laser scanners or image recognition systems, and solved the problem of multi-device synchronization and installation alignment; it can complete the axle type judgment the instant the vehicle passes through the weighing platform, providing real-time data support for weighing and charging.
[0025] (4) Real-time monitoring and predictive maintenance of structural health status: By setting up three-level alarms and ancient-style labels, early faults such as loose bolts and structural fatigue cracks can be detected in advance. Timely warning and maintenance can extend the service life of the weighing platform. Attached Figure Description
[0026] Figure 1 This is a schematic diagram of the structure of the intelligent weighing sensor of the present invention;
[0027] Figure 2 This is a schematic block diagram of the signal processing module of the present invention;
[0028] Figure 3 This is a flowchart of the data processing module of the present invention;
[0029] Figure 4 This is a flowchart of the shaft type recognition module of the present invention. Detailed Implementation
[0030] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.
[0031] like Figure 1 As shown, the present invention provides an intelligent weighing sensor based on a dual-plane beam and high-frequency non-contact vibration measurement; the core components of the sensor include a dual-plane beam 1, a strain sensing unit 2, a vibration sensing unit 3, a temperature sensing unit 4, a signal processing module 5, a data processing module 6, and a shaft type recognition module 7.
[0032] The dual-plane beam 1 is the elastic sensing element of the sensor, bearing the load and generating strain and vibration through its own elastic deformation. The beam is designed with a dual-plane structure to optimize its mechanical properties, improve load sensitivity, and suppress coupled vibrations. When a vehicle passes over it, the beam undergoes slight deformation under stress, simultaneously generating vibrations of different frequencies. This embodiment is manufactured using alloy steel / stainless steel materials, with a rated load range covering 10~40t and a sealing rating of IP68, meeting the requirements for long-term stable operation in complex industrial environments.
[0033] Strain sensing unit 2 is positioned in the critical strain region of the biplane beam 1 to acquire strain signals directly caused by the load. The strain sensing unit 2 employs a high-precision resistance strain gauge with a sensitivity of 1.5 mV / V (350Ω bridge). This resistance strain gauge is precisely attached to the neutral layer surface of the biplane beam 1. The neutral layer is the region of minimum stress when the beam undergoes bending deformation; attaching the strain gauge here allows for more accurate capture of the beam's pure load strain, reducing interference from other factors. The strain acquired by the strain gauge is converted into a mass signal using Hooke's law, thereby enabling weighing.
[0034] The vibration sensing unit 3 is non-contactly positioned near the elastic deformation region of the biplane beam 1. Non-contact measurement avoids the wear, fatigue, and signal interference problems associated with direct contact with the beam. This vibration sensing unit 3 acquires vibration signals from the biplane beam 1 at a sampling frequency greater than 100 kHz. The high sampling frequency captures rich high-frequency vibration modes generated by the beam under load impact, containing structural health information and shaft shape identification clues that traditional low-frequency sensors cannot obtain. The vibration sensing unit 3 employs a nanometer time-grating module. The nanometer time-grating module measures the micro-displacement changes of the biplane beam non-contactly through electromagnetic induction; it has the advantages of high sensitivity, high bandwidth, and non-contact operation, enabling precise capture of the beam's minute vibrations.
[0035] Temperature sensing unit 4 is mounted on the double-plane beam 1 to collect ambient temperature data in real time. Since temperature significantly affects the performance of both the strain sensor and the elastomer material, the data collected by temperature sensing unit 4 will be used for subsequent temperature compensation to ensure the accuracy of the weighing results. Temperature sensing unit 4 uses a PT1000 platinum resistance temperature sensor with an accuracy of ±0.1℃, providing data support for environmental compensation.
[0036] Signal processing module 5 is electrically connected to strain sensing unit 2, vibration sensing unit 3, and temperature sensing unit 4. Its main task is to receive and perform preliminary processing on the raw strain signal, high-frequency vibration signal, and temperature data to generate force measurement data, vibration characteristic data, and temperature compensation data.
[0037] like Figure 2As shown, the signal processing module 5 includes a first processing channel, a second processing channel, and a temperature processing unit. The first processing channel processes the strain signal from the strain sensing unit 2 and mainly includes a Wheatstone bridge, an instrumentation amplifier, a low-pass filter, and an analog-to-digital converter (ADC). The instrumentation amplifier amplifies the weak strain signal; the low-pass filter filters out high-frequency noise, retaining the effective strain signal; and the ADC converts the analog strain signal into a digital signal. The second processing channel processes the high-frequency vibration signal from the vibration sensing unit 3 and mainly includes a charge amplifier, a high-pass filter, and a high-speed analog-to-digital converter (ADC). The charge amplifier amplifies the weak charge signal output by the vibration sensor; the high-pass filter filters out low-frequency interference, especially low-frequency vibrations close to the strain signal; and the high-speed ADC converts the amplified high-frequency vibration signal into a digital signal at a sampling frequency greater than 100kHz. To adapt to different loads and vibration amplitudes, both the first and second processing channels include adaptive gain adjustment circuits to optimize the signal-to-noise ratio and measurement accuracy.
[0038] Data processing module 6 is communicatively connected to signal processing module 5. It receives force measurement data, vibration characteristic data, and temperature compensation data generated by signal processing module 5. Data processing module 6 includes a feature extraction unit, an LSTM neural network model, a health status assessment and alarm unit, and a fault diagnosis unit. The feature extraction unit extracts time-domain features, frequency-domain features, and time-series features from the vibration characteristic data to form a multi-dimensional feature vector. These features may include root mean square, peak value, kurtosis factor, harmonic energy, spectral entropy, wavelet coefficients, etc. Subsequently, data processing module 6 inputs the force measurement data, multi-dimensional feature vector, and temperature compensation data into an LSTM-based neural network model for deep processing. This LSTM model is pre-trained and can learn complex mapping relationships from multi-source heterogeneous data, ultimately outputting high-quality quality measurement values, structural health index values, and fault labels. The health status assessment and alarm unit evaluates the current health status of the equipment based on the structural health index and triggers alarms of different levels. The fault diagnosis unit identifies the fault type based on the fault label. The fault types include, but are not limited to: strain gauge drift, vibration signal loss, temperature sensor failure, ADC conversion abnormality, power fluctuation, communication interruption, structural fatigue, and bolt loosening.
[0039] Data processing module 6 receives force measurement data, vibration characteristic data, and temperature compensation data from signal processing module 5. First, the raw vibration characteristic data is analyzed in depth in the feature extraction unit. Time-domain features include, but are not limited to, statistical indicators such as mean, variance, peak value, peak-to-peak value, root mean square, kurtosis, margin, and impulse factor. These indicators reflect the instantaneous energy and impact characteristics of the vibration signal. Frequency-domain features are obtained through methods such as Fourier transform or wavelet transform, including the dominant frequency, the energy ratio of harmonic components, bandwidth, and spectral entropy. These features can reveal potential problems such as structural resonance, wear, and loosening. Time-series features utilize autocorrelation functions, cross-correlation functions, and higher-order statistics to capture the dynamic patterns of vibration signal changes over time. After integration, a rich multidimensional feature vector with 128 dimensions is formed. This 128-dimensional multidimensional feature vector, along with the force measurement data and temperature compensation data, is input into an LSTM-based neural network model. The LSTM-based neural network model is a three-layer bidirectional LSTM structure. Bidirectional LSTM can simultaneously learn the forward and reverse dependencies of sequence data, thus capturing the temporal dynamics of vibration signals more comprehensively. Specifically, the input layer receives a 128-dimensional feature vector (containing vibration features, force measurements, and temperature compensation data), and the hidden layers are configured with 64 neurons, 128 neurons, and 64 neurons respectively. These three hidden layers progressively extract and abstract higher-level features. The output layer outputs a three-dimensional vector, including mass measurements, structural health index values, and fault labels.
[0040] The mass measurement values described in this embodiment not only consider mechanical signals but also incorporate dynamic information contained in vibration signals and the influence of temperature on sensor response, thereby achieving more accurate and robust mass measurement. This compensation mechanism can effectively suppress the impact of vehicle dynamic loads, vibrations, and changes in ambient temperature on the accuracy of mass measurement.
[0041] The health status assessment and alarm unit receives the structural health index value output by the LSTM model. Based on this index, the system can evaluate the health status of the sensor and trigger alarms of the corresponding level. The sensor adopts a three-level alarm mechanism: (1) When the structural health index is lower than the first threshold, a first-level warning is triggered, indicating that there may be a slight abnormality or performance degradation. (2) When the structural health index is lower than the second threshold, a second-level warning is triggered, indicating that there is a moderate abnormality or potential fault risk, and it is recommended to check. (3) When the structural health index is lower than the third threshold, a third-level warning is triggered and a fault label is output, indicating that the sensor's working status is seriously abnormal or has failed, and it needs to be stopped immediately for inspection or replacement.
[0042] The structural health index is derived through a comprehensive assessment using techniques such as vibration spectrum analysis. It reflects the health status of the beam structure (such as loose bolts, fatigue crack propagation, and changes in elastic body stiffness) by monitoring the drift of specific resonant frequencies, changes in damping ratios, or anomalies in high-frequency vibration modes.
[0043] The fault diagnosis unit receives fault labels output by the LSTM model. These fault labels are used to identify the types of sensor hardware faults. These fault types can include: strain gauge drift (such as zero-point drift or sensitivity changes), vibration signal loss (such as sensor disconnection), temperature sensor failure (such as abnormal readings), ADC conversion abnormalities (such as excessive quantization error), power fluctuations, communication interruptions, structural fatigue (such as material fatigue in a biplane beam), and loose bolts (affecting the beam's natural frequency and damping). By outputting accurate fault labels, the system can quickly locate problems and guide maintenance personnel to perform precise repairs.
[0044] The axle type identification module 7 uses the vibration displacement signal collected by the high-frequency vibration sensing unit 3 to identify the vehicle's axle type. Specifically, it analyzes the time-domain waveform of the vibration displacement signal and calculates the vehicle's wheelbase based on the time interval between peaks (the wheelbase varies for different axle types). Simultaneously, it extracts the spectral energy characteristics of the vibration displacement signal in a specific frequency band (e.g., 20-50Hz), which is typically related to the resonance characteristics of the vehicle's tires, suspension system, and different axle types. Based on the calculated wheelbase and the extracted spectral energy characteristics, the axle type identification module 7 uses a built-in classification model (which can be a machine learning or deep learning model) to determine the accurate axle type of the passing vehicle.
[0045] The axle type identification module 7 performs time-domain analysis on the raw vibration displacement signal collected by the vibration sensing unit 3. By accurately identifying wave peaks and calculating the time interval between adjacent wave peaks, it can estimate the impact time caused by different axles passing by, and thus deduce the vehicle's wheelbase information. Simultaneously, it performs spectral analysis on the raw vibration displacement signal, paying particular attention to the low-frequency band of 20-50Hz, as the vehicle's suspension system, tire-road contact, and the weight characteristics of different axle types often generate significant vibration energy in this band. For example, the drive axle, due to its greater mass and connection to the engine transmission system, may generate a stronger vibration response at a specific frequency when it passes by. Through comprehensive analysis of the wheelbase and the spectral energy characteristics of this specific frequency band, combined with a pre-trained classification model (such as support vector machine, random forest, or small neural network), it can accurately determine the vehicle's axle type, such as single-axle, dual-axle, or triple-axle, and even whether it is a drive axle or a non-drive axle.
[0046] This embodiment describes an intelligent weighing method based on the above-mentioned intelligent weighing sensor and its application in a dynamic truck scale on a highway.
[0047] Reference Figure 4 This invention provides an intelligent weighing method based on a dual-plane beam and high-frequency non-contact vibration measurement. This method utilizes the intelligent weighing sensor described in any one of claims 1 to 7, and specifically includes the following steps:
[0048] Step 1: Data Synchronization and Acquisition.
[0049] As the vehicle passes over the double-plane beam 1, strain sensing unit 2 synchronously acquires strain signals caused by the load; vibration sensing unit 3 synchronously acquires minute vibration signals of the double-plane beam 1 in a non-contact manner at a sampling frequency greater than 100kHz; simultaneously, temperature sensing unit 4 synchronously acquires the current ambient temperature data. All data is precisely timestamped.
[0050] Step 2: Parallel signal processing and feature data generation.
[0051] Signal processing module 5 performs parallel preprocessing on the strain signal, vibration signal, and temperature data acquired in step 1. The first processing channel amplifies, low-pass filters, and performs analog-to-digital conversion on the strain signal to generate raw force measurement data. The second processing channel amplifies, high-pass filters, and performs high-speed analog-to-digital conversion on the strain vibration signal to generate raw vibration data. Simultaneously, the temperature sensor data undergoes preliminary processing to form temperature-compensated data. During this processing, an adaptive gain adjustment circuit ensures signal quality. Finally, the signal processing module generates force measurement data, vibration characteristic data containing time and frequency domain information, and temperature-compensated data.
[0052] Step 3: Multidimensional feature extraction and intelligent model inference.
[0053] Data processing module 6 receives the force measurement data, vibration feature data, and temperature compensation data generated in step 2. Next, the feature extraction module extracts rich time-domain features (such as root mean square, peak value, and kurtosis), frequency-domain features (such as spectral energy distribution and dominant frequency), and time-series features from the vibration feature data. These features are integrated with the force measurement data and temperature compensation data to form a high-dimensional (e.g., 128-dimensional) multidimensional feature vector. Subsequently, this multidimensional feature vector is input into a pre-trained LSTM-based neural network model. The LSTM model performs inference and outputs three key results: high-precision mass measurement values, structural health index values reflecting the sensor and structural condition, and fault labels for fault diagnosis.
[0054] Step 4: Health status assessment and alarm.
[0055] The health status assessment and alarm module evaluates the health status of the sensors based on the structural health index value output in step 3. The system presets three alarm thresholds. If the structural health index is lower than the first threshold, a level 1 alarm is triggered; if it is lower than the second threshold, a level 2 alarm is triggered; if it is lower than the third threshold, a level 3 alarm is triggered, and a specific fault label is output simultaneously, indicating possible structural fatigue, loose bolts, or sensor hardware failure.
[0056] Step 5: Axis type identification.
[0057] The axle type identification module 7 operates independently, performing in-depth analysis of the vibration displacement signal acquired in step 1. First, it analyzes the time-domain waveform of the vibration signal, identifying the shock wave peaks generated when the vehicle axle passes by, and calculating the vehicle's wheelbase based on the time interval between the peaks. Next, it extracts the spectral energy characteristics of the vibration displacement signal in the 20-50Hz frequency band. Finally, based on the calculated wheelbase and spectral energy characteristics, it uses a pre-trained classification model to determine and output the vehicle's axle type information.
[0058] This intelligent weighing method can be used in overload detection systems for dynamic truck scales on highways. In this application scenario, the invention can simultaneously output the vehicle's total weight (obtained by accumulating axle loads), axle loads (mass measurements of each axle as it passes), accurate axle type identification information, structural health status information (health index and warnings), and detailed fault diagnosis information. This marks a significant advancement in dynamic weighing systems, moving from simple mass measurement to a multi-dimensional, intelligent, comprehensive sensing system, greatly improving the system's functionality, reliability, and maintenance efficiency.
[0059] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.
Claims
1. An intelligent load cell based on a two-plane beam and high-frequency non-contact vibration measurement, characterized by, The application relates to a vehicle axle load measuring device, which comprises the following components: a double-plane beam body for bearing load and generating strain and vibration; a strain sensing unit arranged on the double-plane beam body for collecting strain signals; a vibration sensing unit arranged non-contactly near an elastic deformation area of the double-plane beam body for collecting vibration signals at a sampling frequency greater than 100 kHz; a temperature sensing unit arranged on the double-plane beam body for collecting environmental temperature data; a signal processing module electrically connected with the strain sensing unit, the vibration sensing unit and the temperature sensing unit for receiving and processing the strain signals, the vibration signals and the temperature data to generate force measurement data, vibration characteristic data and temperature compensation data; a data processing module in communication connection with the signal processing module for receiving the force measurement data, the vibration characteristic data and the temperature compensation data, extracting time-domain characteristics, frequency-domain characteristics and time sequence characteristics from the vibration characteristic data to form a multi-dimensional feature vector; inputting the force measurement data, the multi-dimensional feature vector and the temperature compensation data into an LSTM-based neural network model and outputting mass measurement values, structure health index values and fault labels; evaluating the health state according to the structure health index and triggering different levels of alarms; an axle type identification module for analyzing the time-domain waveform of the vibration displacement signals, calculating the wheelbase of a vehicle according to the wave peak time interval, extracting the frequency spectrum energy characteristics of the vibration displacement signals in the 20-50 Hz frequency band and judging the axle type through a classification model based on the wheelbase and the frequency spectrum energy characteristics.
2. The smart load sensor based on dual plane beam and high frequency non-contact vibration measurement according to claim 1, characterized in that, The strain sensing unit adopts high-precision resistance strain gauges attached to the surface of the neutral layer of the double-plane beam body.
3. The smart load sensor based on dual plane beam and high frequency non-contact vibration measurement according to claim 1, characterized in that, The vibration sensing unit adopts a nano time grating module to non-contactly measure the micro displacement changes of the double-plane beam body through electromagnetic induction principle.
4. The smart load sensor based on dual plane beam and high frequency non-contact vibration measurement according to claim 1, characterized in that, The signal processing module comprises: a first processing channel for instrument amplification, low-pass filtering and analog-digital conversion of the strain signals; a second processing channel for charge amplification, high-pass filtering and high-speed analog-digital conversion of the vibration signals; wherein the first processing channel and the second processing channel both contain adaptive gain adjustment circuits.
5. The smart load sensor based on dual plane beam and high frequency non-contact vibration measurement according to claim 1, characterized in that, The LSTM-based neural network model is a three-layer bidirectional LSTM structure, the input layer receives a 128-dimensional feature vector, the hidden layers are configured with 64 / 128 / 64 neurons in sequence, and the output layer outputs a three-dimensional vector of mass measurement values, structure health index values and fault labels.
6. The smart load sensor based on dual plane beam and high frequency non-contact vibration measurement according to claim 1, characterized in that, According to the structure health index, a three-level alarm mechanism is triggered: when the structure health index is lower than a first threshold value, a first-level early warning is triggered; when the structure health index is lower than a second threshold value, a second-level early warning is triggered; when the structure health index is lower than a third threshold value, a third-level early warning is triggered and a fault label is outputted. The mass measurement values are compensated mass values calculated through a multi-physical field coupling compensation mechanism; the structure health state evaluation results include a structure health index or wear early warning outputted based on vibration spectrum analysis; the fault diagnosis information contains a fault label of a fault type code.
7. The smart load sensor based on dual plane beam and high frequency non-contact vibration measurement according to claim 1, characterized in that, The fault label is used to identify the sensor hardware fault type, which includes strain gauge drift, vibration signal loss, temperature sensor failure, ADC conversion anomaly, power fluctuation, communication interruption, structure fatigue and bolt loosening.
8. An intelligent weighing method based on a two-plane beam and high-frequency non-contact vibration measurement, characterized by, The method comprises the following steps: Step 1: synchronously collecting strain signals caused by load through the strain sensing unit, synchronously collecting vibration signals of the double-plane beam body in a non-contact manner through the vibration sensing unit, and synchronously collecting environmental temperature data through the temperature sensing unit; Step 2: performing parallel processing on the strain signals, vibration signals and temperature data to respectively generate force measurement data, vibration characteristic data and temperature compensation data; Step 3: extracting a multi-dimensional feature vector from the vibration characteristic data, and inputting the force measurement data, multi-dimensional feature vector and temperature compensation data into a neural network model based on LSTM to output mass measurement values, structure health index values and fault labels; Step 4: evaluating the health state according to the structure health index values and triggering corresponding level alarms; Step 5: analyzing time-domain waveforms and frequency-domain characteristics of the vibration signals through an axle type identification module to identify vehicle axle types.
9. The smart weighing method based on the two-plane beam and high-frequency non-contact vibration measurement according to claim 8, characterized in that, The intelligent weighing method is used in an over-limit detection system of a highway dynamic truck scale, and can synchronously output vehicle total weight, axle weight, axle type identification information, structure health state information and fault diagnosis information.