A bearing load measuring device and a data acquisition and transmission method thereof

Abstract

This invention belongs to the field of ship shafting condition monitoring and fault diagnosis technology, specifically relating to a bearing load measurement device and its data acquisition and transmission method. This invention effectively suppresses temperature drift, mechanical vibration, and electromagnetic interference by rationally arranging strain gauges, employing high-precision synchronous sampling, and combined denoising processing, thereby improving the accuracy of signal acquisition. Embedding moment balance and beam strain consistency equations into the machine learning model significantly improves the accuracy of load prediction. High-speed multi-channel synchronous sampling technology ensures data time alignment, and a multi-algorithm denoising scheme combining wavelet thresholding and bandpass filtering effectively filters out mechanical vibration and electromagnetic interference. A multi-mode communication module enables separate channel transmission of real-time load data and raw strain data, balancing the low-latency requirements of online monitoring with the large-scale data archiving requirements of offline analysis. The introduction of edge computing significantly reduces cloud transmission pressure and improves the real-time performance of load measurement.

Description

Technical Field

[0001] This invention belongs to the field of ship shafting condition monitoring and fault diagnosis technology, specifically relating to a bearing load measuring device and its data acquisition and transmission method. Background Technology

[0002] The coupling between the ship's shafting and hull causes drastic changes in the relative position and load of the bearings and drive shafts. The unsteady interaction forces in the rotating bearing-shaft system are characterized by nonlinearity, high frequency, and large amplitude, often exceeding the measurement range and frequency response bandwidth of traditional measuring equipment, making measurement extremely difficult. Real-time acquisition of the dynamic load in the rotating bearing-shaft system is fundamental for judging the shafting system's operating condition.

[0003] Existing bearing load calculation methods largely rely on sophisticated finite element analysis or empirical formulas. These methods are not only complex in modeling and computationally inefficient, but also ill-suited for complex conditions such as high-speed impacts, exhibiting poor extrapolation performance and failing to meet practical monitoring requirements. Furthermore, traditional measuring equipment suffers from isolated sensor acquisition, data processing, and transmission, making signals susceptible to mechanical vibration and electromagnetic interference. The lack of high-speed multi-channel synchronous sampling capabilities and poor data time alignment further reduce accuracy. In addition, traditional measuring devices have limited power supply systems, making them unsuitable for applications without external power sources in drive shaft systems. Sensor and acquisition node installation stability is insufficient, and communication methods are limited, failing to balance low-latency real-time data transmission with batch archiving of raw data, thus hindering the dual requirements of online monitoring and offline model optimization. Therefore, there is an urgent need to develop an integrated and intelligent bearing load measurement device and corresponding data acquisition and transmission methods to achieve real-time, accurate, and reliable measurement and transmission of bearing dynamic loads. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to provide a bearing load measuring device and its data acquisition and transmission method, which can realize high-precision, real-time monitoring and efficient and reliable transmission of dynamic load of the main drive shaft bearing of a ship, and meet the actual needs of ship shafting condition monitoring.

[0005] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is as follows: I. A bearing load measuring device This invention provides a bearing load measuring device, mainly comprising: a strain gauge 1, a power supply module 2, a data acquisition module 3, a calculation module 4, and a transmission module 5; the strain gauge 1 is used to acquire the original analog load strain signal during bearing operation in real time; the data acquisition module 3 is electrically connected to the strain gauge 1 and is used to filter, amplify, and perform A / D conversion on the original analog signal; the calculation module 4 is an edge computing unit based on an MCU / FPGA architecture and has a built-in pre-trained lightweight machine learning model; the transmission module 5 is a multi-mode communication module and is equipped with dual data channels, used to transmit locally calculated real-time load data and unprocessed original strain data respectively; the power supply module 2 is used to provide a stable voltage to the strain gauge 1, the data acquisition module 3, the calculation module 4, and the transmission module 5.

[0006] Preferably, the strain gauge 1 includes multiple strain detection gauges and corresponding temperature compensation gauges arranged on the surface of the transmission shaft near the bearing to be tested. They are connected by a bridge circuit, and the strain gauge 1 together with the selected shaft acceleration, temperature and speed sensors form a multi-parameter sensing array.

[0007] Preferably, the acquisition module 3 includes a filtering component, an amplification component, and a high-speed multi-channel synchronous sampling ADC component, used to convert the real-time load strain raw analog signal into a strictly time-aligned digital strain sequence, and to establish a unified timestamp for the digital strain sequence.

[0008] Preferably, the calculation module 4 is used to receive the digital strain sequence sent by the acquisition module 3, and at the same time determine whether to perform local edge computing based on system resources, battery balance or host computer instructions. If it is performed, the real-time calculation of the bearing dynamic load is completed through denoising, feature extraction and model inference, and the relative quantity output by the model is converted into an engineering quantity.

[0009] Preferably, the power supply module 2 includes a concave outer shell, a charge / discharge energy storage battery, an energy harvesting unit, and a power monitoring component; the energy harvesting unit is an electromagnetic induction or piezoelectric structure, and the power monitoring component is used to monitor the SOC value in real time.

[0010] Preferably, the measuring device is further equipped with a wireless strain acquisition node and a strain signal receiving device. The wireless strain acquisition node adopts a concave shell design and is attached to the transmission shaft section with special tape at a distance of 3-4 mm from the strain gauge. It is also equipped with a protective film on the outside and has a built-in remote control module, strain acquisition sub-module and wireless signal transmission sub-module.

[0011] Preferably, the strain signal receiving device includes an adapter, a network cable interface, a receiving antenna, a wireless signal receiving submodule, and a wireless signal processing submodule, used to receive data sent by the wireless strain acquisition node and transmit it to a host computer or data monitoring terminal.

[0012] II. A Data Acquisition and Transmission Method for a Bearing Load Measurement Device Based on the same inventive concept, the present invention also provides a data acquisition and transmission method for the bearing load measuring device described above, comprising the following steps: S1, Data Acquisition: Synchronously excite multiple strain gauges and corresponding temperature compensation gauges installed on the surface of the transmission shaft, output the original simulated signal of bearing load strain using the bridge circuit, and select auxiliary sensors to collect the operating parameters of the bearing. S2, Signal Synchronization: The analog signal channels of the strain gauge and the auxiliary sensor are connected to the high-speed multi-channel synchronous sampling ADC of the acquisition module. Multi-channel synchronous sampling is achieved through hardware synchronous triggering, and a unified timestamp is added to the sampled data. S3, Calculation and Judgment: Based on system resources, battery balance, and host computer instructions, determine whether to perform edge computing locally. If yes, proceed to steps S4~S9 to perform local edge computing and output real-time load data after completion. If no, proceed directly to step S10 to retain the original data and upload it to the host computer or cloud for offline analysis. S4, Denoising: Wavelet thresholding and IIR / Biquad bandpass filtering are combined to suppress noise in the synchronized digital strain sequence in order to obtain a denoised and clean strain sequence. S5, Feature Extraction: Extract multidimensional features from the denoised strain sequence within a preset sliding time window, and perform Z-Score normalization on the extracted multidimensional features. S6, Input feature data: Load the normalized feature vector into the model input tensor according to the preset format, and at the same time concatenate the working condition parameters into the model input tensor according to the channel dimension to improve the robustness of the model; S7, Machine Learning Algorithm: Calls a pre-trained lightweight model to perform inference operations on the MCU / FPGA architecture of the computing module and outputs the predicted values ​​of the radial / axial load of the bearing; S8, Bearing load calculation: Combining the geometric parameters of the drive shaft with the calibration coefficients, the relative quantities output by the model are converted into engineering quantities, and a real-time load data frame with confidence level is generated. S9, Dynamic load data encapsulation: The valid real-time load data frame output in step S8 is encapsulated according to the industrial communication protocol, and a data check bit is added to obtain dynamic load data that can be directly transmitted. S10, Original strain data retention: The system retains the original strain data and auxiliary parameter data that were synchronously sampled in step S2 but have not been processed, for post-processing on the host computer or in the cloud. S11, Data Transmission: A multi-mode communication module is used to continuously report the dynamic load data to the monitoring terminal via low-latency UDP / BLE broadcast or CAN message for online alarm and control; simultaneously, the raw strain data is compressed in segments and transmitted via Wi-Fi. Batch uploads can be made to the server via Fi or wired interface for historical archiving and model retraining.

[0013] Preferably, in step S3, determining whether to perform edge computing locally specifically includes: Condition 1: If CPU utilization is <80%, memory remaining is >30%, SOC is ≥20%, or a local computing instruction is received from the host computer, then proceed to steps S4~S9 to execute local edge computing; Condition 2: If condition 1 is not met or an analysis command is received from the host computer cloud, skip the local calculation and directly proceed to step S10 to retain the original data.

[0014] Preferably, in step S8, the conversion formula for the quantity F is as follows: F = J × f(ε) + C In the formula, J is the calibration coefficient, f(ε) is the relative quantity of the model output, and C is the correction value; J and C are both obtained from the bearing static calibration test and are corrected in combination with the geometric parameters of the transmission shaft during calculation; and the data is determined to be valid and output only when the confidence level of the real-time load data frame is ≥90%.

[0015] Compared with the prior art, the present invention has the following main advantages: 1. This invention enables high-precision, real-time measurement of dynamic loads on high-speed transmission shaft bearings. Through reasonable strain gauge arrangement, high-precision synchronous sampling, and combined noise reduction processing, it effectively suppresses temperature drift, mechanical vibration, and electromagnetic interference, thereby improving the accuracy of signal acquisition. By embedding torque balance and beam strain consistency equations into the machine learning model, the accuracy of load prediction is significantly improved, making it suitable for nonlinear, high-frequency, and large-amplitude ship shaft load conditions.

[0016] 2. This invention employs high-speed multi-channel synchronous sampling technology to ensure data time alignment. Combined with wavelet threshold denoising and bandpass filtering, a multi-algorithm denoising scheme effectively filters out mechanical vibration and electromagnetic interference, improving data acquisition accuracy. The multi-mode communication module enables separate channel transmission of real-time load data and raw strain data, balancing the low latency requirements of online monitoring with the large data archiving requirements of offline analysis. The introduction of edge computing significantly reduces cloud transmission pressure and improves the real-time performance of load measurement. Attached Figure Description

[0017] Figure 1 This is an overall schematic diagram of the bearing load measuring device in an embodiment of the present invention; Figure 2 This is a flowchart of the data acquisition and transmission method of the bearing load measuring device in an embodiment of the present invention.

[0018] In the diagram: 1-Strain gauge, 2-Power supply module, 3-Acquisition module, 4-Calculation module, 5-Transmission module. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention. Furthermore, the technical features involved in the various embodiments of this invention described below can be combined with each other as long as they do not conflict with each other.

[0020] It should be noted that, depending on the implementation needs, the various steps / components described in this application can be broken down into more steps / components, or two or more steps / components or parts of the operation of steps / components can be combined into new steps / components to achieve the purpose of this invention.

[0021] In this invention, unless otherwise expressly specified and limited, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this invention, "a plurality of" means at least two, such as two, three, etc., unless otherwise expressly and specifically limited.

[0022] Example 1: This example provides a bearing load measuring device, such as... Figure 1 As shown, it mainly includes: 1) Strain gauges: These include multiple sets of strain gauges and temperature compensation gauges placed on the shaft surface near the bearing. They are connected by a bridge circuit and are used to acquire the original simulated load strain signal during bearing operation in real time. The temperature compensation gauge is used to eliminate the interference of ambient temperature changes on strain detection and ensure the accuracy of the acquired signal.

[0023] 2) Power Supply Module: Provides stable power to all modules of the device, including the concave housing, rechargeable and discharging energy storage battery, energy harvesting unit, and power monitoring component; the concave housing adapts to the axial surface structure to ensure stable installation; the rechargeable and discharging energy storage battery can be charged via an external power source through a USB interface, and can work continuously for no less than 8 hours after being fully charged; the energy harvesting unit adopts an electromagnetic induction or piezoelectric structure, and uses the mechanical energy of the shaft rotation to achieve self-powering; the power monitoring component monitors the battery SOC value in real time, and triggers the system's low-power operation strategy when the SOC is lower than the preset threshold.

[0024] 3) Acquisition module: Electrically connected to strain gauge 1, including filtering components, amplification components and high-speed multi-channel synchronous sampling ADC components, used to receive the original analog signal from strain gauge 1, and sequentially perform signal filtering, amplification and A / D conversion, converting the analog signal into a strictly time-aligned digital strain sequence, and establishing a unified timestamp for the digital sequence.

[0025] 4) Calculation module: This is an edge computing unit based on MCU / FPGA, with a built-in pre-trained lightweight machine learning model and a database of bearing geometric parameters and calibration coefficients; it is used to receive the digital strain sequence of the acquisition module 3, and determine whether to perform local edge computing based on system resources, battery balance or host computer instructions. If it is performed, it completes the real-time measurement of the bearing dynamic load through denoising, feature extraction and model inference, and converts the relative quantity into an engineering quantity on the order of N / kN.

[0026] 5) Transmission module: It is a BLE+Wi-Fi+CANBUS multi-mode communication module, which is electrically connected to the calculation module 4. It is equipped with dual data channels, which are used to transmit the real-time load data after local calculation and the raw strain data before processing.

[0027] In this embodiment, the device is also equipped with a wireless strain acquisition node and a strain signal receiving device. The wireless strain acquisition node has a concave shell design and is attached to the shaft section 3-4 mm away from the strain gauge using special tape. It is externally protected and has a built-in remote control module, strain acquisition submodule, and wireless signal transmission submodule. The strain signal receiving device includes an adapter, a network cable interface, a receiving antenna, a wireless signal receiving submodule, and a wireless signal processing submodule, which are used to receive wirelessly transmitted data and import it into a host computer or data monitoring terminal.

[0028] Example 2: This example provides a bearing load measuring device that monitors the load strain signal of the bearing in real time during operation by fixing a strain gauge 1 near the bearing to be tested.

[0029] In this embodiment, the analog signal generated by the strain gauge 1 is transmitted to the acquisition module 3. The acquisition module 3 filters, amplifies, and performs A / D conversion on the analog signal to convert it into a digital signal. The digital signal is input to the calculation module 4, which analyzes and calculates the acquired digital signal according to a preset algorithm to obtain bearing load data. The load data output by the calculation module is transmitted to a host computer or data monitoring terminal via a wireless or wired connection through the transmission module 5 for real-time monitoring and analysis. The power supply module 2 provides a stable voltage to the strain gauge 1, acquisition module 3, calculation module 4, and transmission module 5 to ensure the normal operation of each module.

[0030] In this embodiment, strain signal transmission is accomplished through a wireless strain acquisition node. The wireless strain acquisition node consists of a housing, strain signal pins, power supply pins, a remote control module, a strain acquisition module, and a wireless signal transmission module. The housing features a concave design, allowing for a closer fit to the axial surface during installation, thus ensuring installation stability. During operation, power is supplied through the power supply pins, and the strain signal line is connected to the input pins, which in turn connects to the strain acquisition module. The data is then transmitted to the signal receiver via the wireless signal transmission module.

[0031] In this embodiment, since the signal acquisition node is powered by an external power supply, a power supply module is required to provide power support. The wireless strain gauge power supply module consists of a concave outer shell and an internal rechargeable and discharging energy storage battery. The concave outer shell design conforms to the shaft surface, ensuring stable installation. During operation, the wireless strain gauge acquisition node is powered via a lead wire. After use, it can be charged by connecting to a fixed power source via a USB interface. After a full charge, the wireless strain gauge power supply can work continuously for approximately 8 hours. In terms of power supply assurance, the entire process is powered by a combination of the shaft-built-in battery and the energy acquisition unit (electromagnetic induction or piezoelectric); when the SOC is lower than the threshold, the system automatically disables local calculations, retaining only low-power sampling and data packaging functions.

[0032] In this embodiment, after the wiring with the strain gauge is completed and it is checked that it can be used normally, the strain node device is attached to the strain gauge 3-4mm away from the strain gauge using a specific tape and shaft section, and a protective film is deployed on the outside. In this embodiment, the strain signal receiving device comprises an adapter, a network cable interface, a receiving antenna, a wireless signal receiving module, and a wireless signal processing module. During the experiment, the strain signal receiving device is fixed next to the computer and powered by the adapter. Simultaneously, data is imported into the computer via a fiber optic network cable to complete the data collection process.

[0033] Example 3: Based on the same inventive concept, this example also provides a data acquisition and transmission method for the bearing load measuring device as described above, such as... Figure 2 As shown, it includes the following steps: Step S1, Data Acquisition: Synchronously excite multiple sets (K sets in this example) of strain gauges and temperature compensation gauges mounted on the surface of the transmission shaft, and output the original simulated signal of bearing load strain using a full-bridge circuit; at the same time, select auxiliary sensors such as acceleration sensors, temperature sensors, and speed sensors, and keep the sampling frequency of the auxiliary sensors consistent with that of the strain gauges (≥2kHz) to collect information such as acceleration, temperature, and real-time speed of the shaft system to supplement the bearing operating status parameters.

[0034] Step S2, Signal Synchronization: Connect the analog signal channels of the strain gauge and the auxiliary sensor to the high-speed multi-channel synchronous sampling ADC of the acquisition module. Achieve synchronous sampling of all channels through hardware synchronization triggering. The sampling frequency is ≥2kHz. Add a unified timestamp with microsecond precision to the sampled data to generate a strictly time-aligned digital strain sequence and auxiliary parameter sequence.

[0035] Step S3, Calculation Determination (Determine whether to calculate?): The calculation module obtains the remaining battery level through the SOC detection module, the CPU utilization rate and remaining memory through the system monitoring unit, and simultaneously receives instructions from the host computer / monitoring terminal. It selects the data processing method according to the following determination logic: 1) If CPU utilization is <80%, memory remaining is >30%, SOC is ≥20%, or a local computing instruction is received from the host computer, proceed to steps S4~S9 to perform local edge computing, and output real-time load data after completion; 2) If the above conditions are not met or an analysis command is received from the host computer or cloud, skip the local calculation and proceed directly to step S10 to retain the original data and upload it to the host computer or cloud for offline analysis.

[0036] The decision logic is embedded in the control program of the calculation module, and the parameters can be modified via a host computer.

[0037] The local edge computing consists of steps S4 to S9.

[0038] Step S4, Denoising: Noise suppression is performed on the synchronized digital strain sequence using a combination of wavelet thresholding and IIR / Biquad bandpass filtering. First, the db4 wavelet is used to decompose the digital strain sequence into 5 levels, and the soft thresholding method is used to filter out high-frequency electromagnetic interference noise. Then, IIR / Biquad bandpass filtering is used, with the passband frequency set to 0.1~1000Hz, to filter out low-frequency mechanical vibration noise, resulting in a denoised and clean strain sequence.

[0039] Step S5, Feature Extraction: Set a sliding time window of 0.5~2s, and extract multidimensional features from the denoised strain sequence within the time window, including RMS (root mean square), peak value, kurtosis, waveform factor, envelope spectral energy, time-frequency energy vector, etc.; and perform Z-Score normalization on the extracted multidimensional features, the calculation formula is as follows: x norm = (x μ) / σ In the formula, x norm The normalized feature vector is defined by x, where x is the original feature value, μ is the feature mean, and σ is the feature standard deviation. This eliminates the influence of dimensions and yields the normalized feature vector.

[0040] Step S6, Input feature data: Load the normalized feature vector into the model input tensor according to the K×L dimension (K is the number of strain gauges, L is the time window length). At the same time, the operating condition variables such as the real-time rotational speed of the shaft system, the bearing operating temperature, and the ambient humidity are concatenated into the input tensor according to the channel dimension to improve the model's operating condition adaptability and robustness.

[0041] Step S7, Machine Learning Algorithm: Call a pre-trained lightweight model (such as 1) DCNN+GRU, SVR, or XGBoost performs inference on the heterogeneous architecture of the computing module (MCU / FPGA) and outputs predicted values ​​of radial / axial loads; model weights can be iterated online via OTA upgrades.

[0042] Step S8, Bearing Load Calculation: Combining the transmission shaft geometric parameters and calibration coefficients, the relative quantities output by the model are converted into engineering quantities (N or kN), generating a real-time load data frame with a confidence index; the engineering quantity conversion formula is as follows: F = J × f(ε) + C Where J is the calibration coefficient, f(ε) is the relative quantity of the model output, and C is the correction value; J and C are obtained from the bearing static calibration test and are corrected in combination with the geometric parameters of the transmission shaft (diameter, elastic modulus, Poisson's ratio) during calculation; real-time load data frames are generated, and the data frames contain timestamps, radial load values, axial load values, confidence index, and operating parameters. The data is considered valid and output only when the confidence level is ≥90%.

[0043] Step S9, Dynamic Load Data Encapsulation: The valid real-time load data frame output from S8 is encapsulated according to the industrial communication protocol, and a data check bit is added to prevent data loss or tampering during transmission, resulting in dynamic load data that can be directly transmitted.

[0044] Step S10, Original Strain Data Retention: The system retains the original strain data and auxiliary parameter data that were synchronously sampled in step S2 but have not been processed, for subsequent processing by the host computer or cloud.

[0045] Step S11, Data transmission: Using a multi-mode communication module (BLE+Wi-Fi) Fi+CANBUS enables differentiated data transmission. Real-time load data is continuously reported to the monitoring terminal via low-latency UDP / BLE broadcast or CAN messages for online alarm and control; raw strain data is fragmented, compressed, and then transmitted via Fi+CANBUS. Batch uploads can be made to the server via Fi or wired interface to achieve historical archiving and model retraining.

[0046] In this embodiment, the bearing dynamic load calculation method based on machine learning in steps S6-S8 specifically includes: constructing a time window tensor of a specified length L as network input by collecting time-series data generated by K strain gauges arranged on the shaft; the network backbone adopts a residual one-dimensional convolution structure to extract inertial coupling features; and superimposing force in the loss function. Moment balance residuals and bending moment Strain consistency terms are implemented to improve physical interpretability. The process is as follows: (1) Acquisition steps: Collect the strain measurement sequence of K groups of strain gauges placed on the drive shaft within a preset time window L to form the first input tensor;

[0047] (2) Prediction step: Input the first input tensor into a one-dimensional fully connected hybrid neural network containing residuals and dilated convolutions to obtain a bearing dynamic load prediction vector with six components (or the required components).

[0048] (3) Training steps: Define the load loss function:

[0049] For the calculated and calibrated bearing load data, the minimization of the network model loss function includes a composite loss function with four error terms to obtain the optimal model parameters. The first error term is the mean square error between the predicted load and the calibrated load. The second error term is the moment balance residual calculated based on the predicted load. The third error term is the moment-strain consistency residual calculated based on beam theory. The fourth error term is the L2 regularization of the network parameters.

[0050] (4) Reasoning steps: Use the trained model to predict the load in real time for new strain time series data under unknown working conditions.

[0051] Furthermore, all parts of this application that are not described in detail are the same as or implemented using existing technology.

[0052] In summary: 1. This invention enables high-precision, real-time measurement of dynamic loads on high-speed transmission shaft bearings. Through reasonable strain gauge arrangement, high-precision synchronous sampling, and combined noise reduction processing, it effectively suppresses temperature drift, mechanical vibration, and electromagnetic interference, thereby improving the accuracy of signal acquisition. By embedding torque balance and beam strain consistency equations into the machine learning model, the accuracy of load prediction is significantly improved, making it suitable for nonlinear, high-frequency, and large-amplitude ship shaft load conditions.

[0053] 2. This invention employs high-speed multi-channel synchronous sampling technology to ensure data time alignment. Combined with wavelet threshold denoising and bandpass filtering, a multi-algorithm denoising scheme effectively filters out mechanical vibration and electromagnetic interference, improving data acquisition accuracy. The multi-mode communication module enables separate channel transmission of real-time load data and raw strain data, balancing the low latency requirements of online monitoring with the large data archiving requirements of offline analysis. The introduction of edge computing significantly reduces cloud transmission pressure and improves the real-time performance of load measurement.

[0054] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0055] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0056] Those skilled in the art will readily understand that the above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A bearing load measuring device, characterized in that: The system includes a strain gauge (1), a power supply module (2), an acquisition module (3), a calculation module (4), and a transmission module (5). The strain gauge (1) is used to acquire the original simulated load strain signal during the bearing operation process in real time. The acquisition module (3) is electrically connected to the strain gauge (1) and is used to filter, amplify, and perform A / D conversion on the original simulated signal. The calculation module (4) is an edge computing unit based on an MCU / FPGA architecture and has a built-in pre-trained lightweight machine learning model. The transmission module (5) is a multi-mode communication module and is equipped with dual data channels, which are used to transmit the real-time load data after local calculation and the unprocessed original strain data, respectively. The power supply module (2) is used to provide a stable voltage for the strain gauge (1), the acquisition module (3), the calculation module (4), and the transmission module (5).

2. The bearing load measuring device according to claim 1, characterized in that, The strain gauge (1) includes multiple strain detection gauges and corresponding temperature compensation gauges arranged on the surface of the transmission shaft near the bearing to be tested. They are connected by a bridge circuit. The strain gauge (1) together with the selected shaft acceleration, temperature and speed sensors form a multi-parameter sensing array.

3. The bearing load measuring device according to claim 1, characterized in that, The acquisition module (3) includes a filtering component, an amplification component and a high-speed multi-channel synchronous sampling ADC component, which are used to convert the real-time load strain original analog signal into a strictly time-aligned digital strain sequence and establish a unified timestamp for the digital strain sequence.

4. The bearing load measuring device according to claim 1, characterized in that, The calculation module (4) is used to receive the digital strain sequence sent by the acquisition module (3), and at the same time, it determines whether to perform local edge computing based on system resources, battery balance or host computer instructions. If it is performed, it completes the real-time calculation of bearing dynamic load through denoising, feature extraction and model inference, and converts the relative quantity output by the model into engineering quantity.

5. The bearing load measuring device according to claim 1, characterized in that, The power supply module (2) includes a concave outer shell, a charge-discharge energy storage battery, an energy acquisition unit, and a power monitoring component; the energy acquisition unit is an electromagnetic induction or piezoelectric structure, and the power monitoring component is used to monitor the SOC value in real time.

6. The bearing load measuring device according to claim 1, characterized in that, The measuring device is also equipped with a wireless strain acquisition node and a strain signal receiving device. The wireless strain acquisition node adopts a concave shell design and is attached to the transmission shaft section with special tape at a distance of 3-4 mm from the strain gauge. It is also equipped with a protective film on the outside and has a built-in remote control module, strain acquisition sub-module and wireless signal transmission sub-module.

7. A bearing load measuring device according to claim 6, characterized in that, The strain signal receiving device includes an adapter, a network cable interface, a receiving antenna, a wireless signal receiving submodule, and a wireless signal processing submodule, used to receive data sent by the wireless strain acquisition node and transmit it to a host computer or data monitoring terminal.

8. A data acquisition and transmission method for the bearing load measuring device as described in any one of claims 1 to 7, characterized in that, Includes the following steps: S1, Data Acquisition: Synchronously excite multiple strain gauges and corresponding temperature compensation gauges installed on the surface of the transmission shaft, output the original simulated signal of bearing load strain using the bridge circuit, and select auxiliary sensors to collect the operating parameters of the bearing. S2, Signal Synchronization: The analog signal channels of the strain gauge and the auxiliary sensor are connected to the high-speed multi-channel synchronous sampling ADC of the acquisition module. Multi-channel synchronous sampling is achieved through hardware synchronous triggering, and a unified timestamp is added to the sampled data. S3, Calculation and Judgment: Based on system resources, battery balance, and host computer instructions, determine whether to perform edge computing locally. If yes, proceed to steps S4~S9 to perform local edge computing and output real-time load data after completion. If no, proceed directly to step S10 to retain the original data and upload it to the host computer or cloud for offline analysis. S4, Denoising: Wavelet thresholding and IIR / Biquad bandpass filtering are combined to suppress noise in the synchronized digital strain sequence in order to obtain a denoised and clean strain sequence. S5, Feature Extraction: Extract multidimensional features from the denoised strain sequence within a preset sliding time window, and perform Z-Score normalization on the extracted multidimensional features. S6, Input feature data: Load the normalized feature vector into the model input tensor according to the preset format, and at the same time concatenate the working condition parameters into the model input tensor according to the channel dimension to improve the robustness of the model; S7, Machine Learning Algorithm: Calls a pre-trained lightweight model to perform inference operations on the MCU / FPGA architecture of the computing module and outputs the predicted values ​​of the radial / axial load of the bearing; S8, Bearing load calculation: Combining the geometric parameters of the drive shaft with the calibration coefficients, the relative quantities output by the model are converted into engineering quantities, and a real-time load data frame with confidence level is generated. S9, Dynamic load data encapsulation: The valid real-time load data frame output in step S8 is encapsulated according to the industrial communication protocol, and a data check bit is added to obtain dynamic load data that can be directly transmitted. S10, Original strain data retention: The system retains the original strain data and auxiliary parameter data that were synchronously sampled in step S2 but have not been processed, for post-processing on the host computer or in the cloud. S11, Data Transmission: A multi-mode communication module is used to continuously report the dynamic load data to the monitoring terminal via low-latency UDP / BLE broadcast or CAN message for online alarm and control; simultaneously, the raw strain data is compressed in segments and transmitted via Wi-Fi. Batch uploads can be made to the server via Fi or wired interface for historical archiving and model retraining.

9. The data acquisition and transmission method according to claim 8, characterized in that... In step S3, determining whether to perform edge computing locally specifically includes: Condition 1: If CPU utilization is <80%, memory remaining is >30%, SOC is ≥20%, or a local computing instruction is received from the host computer, then proceed to steps S4~S9 to execute local edge computing; Condition 2: If condition 1 is not met or an analysis command is received from the host computer cloud, skip the local calculation and directly proceed to step S10 to retain the original data.

10. The data acquisition and transmission method according to claim 8, characterized in that... In step S8, the conversion formula for the quantity F is as follows: F = J × f(ε) + C In the formula, J is the calibration coefficient, f(ε) is the relative quantity of the model output, and C is the correction value; J and C are both obtained from the bearing static calibration test and are corrected in combination with the geometric parameters of the transmission shaft during calculation; and the data is determined to be valid and output only when the confidence level of the real-time load data frame is ≥90%.

Images