Culvert pipe operation monitoring method and device, electronic equipment and storage medium
Through the combination of deep learning model and deformation sensor data, the problem of low culvert monitoring accuracy in the existing technology is solved, and the accurate judgment of the culvert state and timely discovery of potential problems are achieved, which extends the service life of the culvert.
Patent Information
- Application Number
- CN202411070684.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-06
- Publication Date
- 2024-11-22
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
The existing culvert operation monitoring method relies on a single stress test and simple deformation threshold comparison, and the monitoring accuracy is not high and potential problems cannot be detected sensitively.
The deep learning model is used to combine deformation sensor data, and pre-process it by obtaining real-time deformation data, input it into the pre-trained deep learning model, calculate the reconstruction error and compare it with the preset threshold to judge the normal or abnormal state of the culvert.
It improves detection accuracy, can promptly detect potential faults and abnormalities of the culvert pipe, extends the service life of the culvert pipe, and improves the reliability of the system.
Smart Images

Figure CN119004319A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of culvert monitoring, and in particular to a culvert operation monitoring method, device, electronic equipment and storage medium. Background Art
[0002] Culvert is a kind of building material used in water conservancy projects, municipal engineering, road engineering and other fields. It usually refers to a pipe buried below the surface. The main function of the culvert is to serve as a component of culverts, drainage systems, conveying systems and other engineering projects. It is used to build drainage systems to ensure that water can be discharged smoothly. At the same time, it has the characteristics of corrosion resistance, high temperature resistance and wear resistance.
[0003] At present, the culvert operation monitoring method mainly relies on a single pressure test and a simple deformation threshold comparison. This method applies pressure inside the culvert and then measures the deformation or pressure response of the culvert. The specific operation is to use a pressure gauge or corresponding sensor to measure the performance of the culvert under a certain pressure and compare these data with the preset thresholds. If the measured pressure or deformation data exceeds these thresholds, it indicates that the culvert may have structural integrity problems, such as cracks, corrosion or other physical damage. In this case, the culvert is judged to be abnormal and requires further inspection or repair.
[0004] The existing technology relies on a single pressure test and a simple deformation threshold comparison, so the monitoring accuracy is not high and it cannot sensitively detect potential problems of the culvert. Summary of the invention
[0005] The present invention provides a culvert operation monitoring method, device, electronic equipment and storage medium to improve the accuracy of detection and timely discover potential faults and abnormalities before culvert problems become serious.
[0006] In a first aspect, in order to solve the above technical problem, the present invention provides a culvert operation monitoring method, which is executed by a processor and includes: Acquire the current deformation data of the culvert to be tested under the real-time pressure detected by the deformation sensor; Preprocessing the current deformation data to obtain a current deformation vector; Inputting the current deformation vector into a pre-trained deep learning model to obtain a reconstruction error; Comparing the reconstruction error with a preset error threshold, when the reconstruction error is less than the error threshold, judging that the culvert is working normally; when the reconstruction error is greater than the error threshold, judging that the culvert is working abnormally; The training process of the deep learning model includes: Acquire a first sample deformation data set and a second sample deformation data set; Preprocessing the first sample deformation data set and the second sample deformation data set to obtain a first sample vector set and a deformed second sample vector set; Inputting the first sample vector set into an initial deep learning model to train the initial deep learning model; The second sample vector set is input into the initial deep learning model, the model is evaluated to obtain a comprehensive evaluation index, and when the comprehensive evaluation index meets the preset evaluation index, a trained deep learning model is obtained.
[0007] Preferably, the step of obtaining the current deformation data of the culvert to be tested under the real-time pressure detected by the deformation sensor comprises: Receiving the coded real-time deformation quantity of the culvert to be tested sent by the deformation sensor, decoding the coded real-time deformation quantity of the culvert to be tested, and obtaining current deformation data corresponding to the culvert to be tested; The deformation sensor detects the real-time amount of the culvert deformation, encodes the real-time amount of the culvert deformation, and transmits the encoded real-time amount of the culvert deformation to the processor.
[0008] Preferably, the real-time pressure is the pressure applied to the top of the culvert pipe, which changes in a sinusoidal or cosine law according to time, with a period of T.
[0009] Preferably, the deformation sensor comprises two strain gauge sensors, which are respectively arranged at the top and bottom of the culvert, and are used to measure the strain value of the top and the hoop stress strain value of the bottom; Among them, the current deformation data is the absolute value average of the strain value at the top of the tube and the strain value at the bottom of the tube.
[0010] Preferably, the preprocessing of the current deformation data to obtain the current deformation vector includes: Construct a plane rectangular coordinate system, with the current deformation data as the vertical axis and time as the horizontal axis, and draw a strain curve of the current deformation data changing with time, where the 0 point of the horizontal axis is the moment when the periodic pressure starts to be applied; Uniformly extracting a number of characteristic points within the 0-T time of the plane rectangular coordinate system, and obtaining current deformation data corresponding to each of the characteristic points based on the strain curve; The current deformation vector is constructed based on all the current deformation data.
[0011] Preferably, the first sample deformation data set and the second sample deformation data set are specifically: The first sample deformation data set is pre-measured from a batch of normally working culverts; The second sample deformation data set is pre-measured from a batch of culverts including normally working culverts and abnormally working culverts, wherein the actual working condition of each culvert has been pre-obtained and marked.
[0012] Preferably, the step of inputting the second sample vector set into an initial deep learning model, evaluating the model, obtaining a comprehensive evaluation index, and obtaining a trained deep learning model when the comprehensive evaluation index meets a preset evaluation index comprises: Inputting the second sample vector set into the initial deep learning model for detection, and obtaining a reconstruction error corresponding to each culvert corresponding to the second sample vector set; Comparing the reconstruction error with an error threshold to determine whether the culvert is working properly; Compare the judgment result with the actual working condition of the culvert, calculate the number of normal culverts judged as normal culverts TP, the number of normal culverts judged as abnormal culverts FN, and the number of abnormal culverts judged as normal culverts FP; Compare the comprehensive evaluation index with the preset standard evaluation index. If the comprehensive evaluation index is greater than the standard evaluation index, it means that the model has completed the feature learning of the normal working culvert, and the trained deep learning model is obtained. If the comprehensive evaluation index is less than the threshold, it means that the model has not completed the feature learning of the normal working culvert, and it is necessary to collect the sample vector set again and input it into the model for training; The calculation process of comprehensive evaluation index includes:
[0013]
[0014]
[0015] Among them, F1 is a comprehensive evaluation indicator, the range of F1 is 0 to 1, 1 represents perfect performance, 0 represents the worst performance, P is the precision rate, and R is the recall rate.
[0016] In a second aspect, the present invention provides a culvert operation monitoring device, comprising: A data acquisition module is used to obtain the current deformation data of the culvert to be tested under the real-time pressure detected by the deformation sensor; A data preprocessing module preprocesses the current deformation data to obtain a current deformation vector; An error calculation module inputs the current deformation vector into a pre-trained deep learning model to obtain a reconstruction error; A condition identification module compares the reconstruction error with a preset error threshold, and when the reconstruction error is less than the error threshold, it is determined that the culvert is working normally; when the reconstruction error is greater than the error threshold, it is determined that the culvert is working abnormally; A model training module obtains a first sample deformation data set and a second sample deformation data set; preprocesses the first sample deformation data set and the second sample deformation data set to obtain a first sample vector set and a deformed second sample vector set; inputs the first sample vector set into an initial deep learning model to train the initial deep learning model; inputs the second sample vector set into the initial deep learning model to evaluate the model to obtain a comprehensive evaluation index, and when the comprehensive evaluation index meets the preset evaluation index, a trained deep learning model is obtained.
[0017] In a third aspect, the present invention further provides an electronic device, comprising a processor, a memory, and a computer program stored in the memory and trained to be executed by the processor, wherein the processor implements any one of the above-mentioned culvert operation monitoring methods when executing the computer program.
[0018] In a fourth aspect, the present invention further provides a computer-readable storage medium, wherein the computer-readable storage medium includes a stored computer program, wherein when the computer program is executed, the device where the computer-readable storage medium is located is controlled to execute any one of the culvert operation monitoring methods described above.
[0019] Compared with the prior art, the present invention has the following beneficial effects: the present invention provides a culvert operation monitoring method, device, electronic device and storage medium, the method is executed by a processor, including obtaining current deformation data of a culvert to be tested under real-time pressure detected by a deformation sensor; preprocessing the current deformation data to obtain a current deformation vector; inputting the current deformation vector into a pre-trained deep learning model to obtain a reconstruction error; comparing the reconstruction error with a preset error threshold, when the reconstruction error is less than the error threshold, judging that the culvert is working normally; when the reconstruction error is greater than the error threshold, judging that the culvert is working abnormally; wherein the training process of the deep learning model includes: obtaining a first sample deformation data set and a second sample deformation data set; The first sample deformation data set and the second sample deformation data set are preprocessed to obtain a first sample vector set and a deformation second sample vector set; the first sample vector set is input into an initial deep learning model to train the initial deep learning model; the second sample vector set is input into the initial deep learning model to evaluate the model to obtain a comprehensive evaluation index, and when the comprehensive evaluation index meets the preset evaluation index, a trained deep learning model is obtained. This method uses a deep learning model to learn the characteristics of the deformation data of the culvert under pressure changes, and can identify data that is inconsistent with normal culvert deformation, thereby improving the accuracy of detection and enabling the timely detection of potential faults and abnormalities before culvert problems become serious. BRIEF DESCRIPTION OF THE DRAWINGS
[0020] Figure 1 is a schematic flow chart of a culvert operation monitoring method provided by an embodiment of the present invention; Figure 2 It is a schematic diagram of the deep learning model training process provided by an embodiment of the present invention; Figure 3 It is a schematic diagram of the structure of a culvert operation monitoring device provided in an embodiment of the present invention. DETAILED DESCRIPTION
[0021] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0022] Reference Figure 1 The first embodiment of the present invention provides a culvert operation monitoring method, which is executed by a processor and includes the following steps: S11, obtaining current deformation data of the culvert to be tested under real-time pressure detected by the deformation sensor; S12, preprocessing the current deformation data to obtain a current deformation vector; S13, inputting the current deformation vector into a pre-trained deep learning model to obtain a reconstruction error; S14, comparing the reconstruction error with a preset error threshold, when the reconstruction error is less than the error threshold, it is determined that the culvert is working normally; when the reconstruction error is greater than the error threshold, it is determined that the culvert is working abnormally.
[0023] In step S11, current deformation data of the culvert to be tested under real-time pressure detected by the deformation sensor is obtained.
[0024] In an optional embodiment, the encoded real-time deformation quantity of the culvert to be tested sent by the deformation sensor is received, the encoded real-time deformation quantity of the culvert to be tested is decoded, and the current deformation data corresponding to the culvert to be tested is obtained; wherein the deformation sensor detects the real-time deformation quantity of the culvert, encodes the real-time deformation quantity of the culvert, and transmits the encoded real-time deformation quantity of the culvert to the processor.
[0025] It should be noted that the real-time pressure is the pressure applied to the top of the culvert pipe, which changes in a sinusoidal or cosine manner according to time, with a period of T.
[0026] It should be noted that the deformation sensors include: strain sensors, which use resistance strain gauges to convert strain into resistance changes; optical fiber three-dimensional shape sensors, which use the distribution of two-dimensional strain in the cross section of the optical fiber along the axial direction to achieve three-dimensional shape reconstruction of the optical fiber; fiber Bragg grating (FBG) sensors, the principle of which is that when the temperature, strain, stress or other physical quantities to be measured around the grating changes, the grating period or core emissivity will change, thereby generating a wavelength shift of the grating Bragg signal; vibrating string sensors, which detect strain by measuring the frequency change of the vibrating string.
[0027] Furthermore, in the selection process of deformation sensors, it is crucial to select the right sensor according to the application scenario. Each sensor has its specific advantages and applicable scenarios. Improper selection may lead to inaccurate measurements or even damage to the sensor. Strain sensors are suitable for situations that require high-precision measurements and long-term stable monitoring. Such sensors are usually installed in key parts of culverts, such as stress concentration areas, to monitor strain caused by external loads, temperature changes, etc. The fiber optic three-dimensional shape sensor is suitable for occasions where it is necessary to obtain the three-dimensional shape changes of the culvert. This sensor can be arranged along the length of the culvert to monitor the deformation of the overall shape of the culvert, especially for complex terrain or situations where the culvert has large deformation; the fiber Bragg grating (FBG) sensor is suitable for occasions where real-time monitoring of multi-point strain and temperature changes is required. The FBG sensor can be arranged with multiple measuring points along the culvert to realize strain and temperature monitoring at different positions of the culvert. It is suitable for long-distance monitoring of culverts and occasions that require multi-parameter monitoring; the vibrating string sensor is suitable for occasions that require long-term monitoring and complex environmental conditions. The vibrating string sensor has good long-term stability and environmental resistance, and is suitable for use in harsh environments such as high humidity, high salt, and strong vibration. It is used to monitor the long-term strain changes of the culvert.
[0028] In an optional embodiment, the coded real-time deformation quantity of the culvert to be tested sent by the deformation sensor is received, and the coded real-time deformation quantity of the culvert to be tested is decoded to obtain the current deformation data corresponding to the culvert to be tested, including: data acquisition, the sensor first collects the data of the physical parameters through its detection unit; signal conditioning, the collected raw data is often an analog signal, which needs to be converted into a digital signal through an analog-to-digital converter (ADC). Analog signal conditioning includes steps such as amplification and filtering to ensure the quality and accuracy of the signal; data encoding, the digital signal needs to be encoded into a specific format for transmission; data transmission, the encoded data is transmitted to the controller through the communication interface; data reception and decoding, after the controller receives the data transmitted by the sensor, it needs to decode and convert the encoded data back to the original digital signal or analog signal for further processing and analysis.
[0029] It should be noted that the encoding methods include: digital encoding, converting the numerical value of an analog signal into a digital code, such as binary or hexadecimal; pulse code modulation (PCM), converting the time waveform of an analog signal into a series of pulses, each pulse representing a specific signal value; modulation coding, modulating a digital signal into a format suitable for wireless or wired transmission, such as frequency modulation (FM) or phase modulation (PM).
[0030] Furthermore, the communication interface includes: serial communication, such as RS-232, RS-485, I2C, SPI, etc., which is suitable for short-distance, low-speed data transmission; parallel communication, such as PCI, USB, etc., which is suitable for high-speed, large-capacity data transmission; wireless communication, such as Wi-Fi, Bluetooth, Zigbee, etc., which is suitable for long-distance, flexible deployment scenarios.
[0031] In order to obtain the current deformation data of the culvert to be tested under the real-time pressure detected by the deformation sensor, two implementation methods are provided below: In an optional embodiment, the processor obtains the current deformation data of the culvert to be tested under the real-time pressure detected by the deformation sensor, including: The processor searches for Bluetooth devices to match and establish a Bluetooth connection with the deformation sensor. In this step, the processor's Bluetooth management software lists all detectable Bluetooth devices. The user needs to select the corresponding deformation sensor from the list and may need to enter a pairing code or go through other security authentication processes to complete the pairing. After the pairing is successful, the processor sends a data acquisition instruction to the deformation sensor through the Bluetooth interface. This instruction is a specific signal or data sequence, which instructs the deformation sensor to start performing the data acquisition task; After receiving the command, the deformation sensor activates its built-in data acquisition system to measure the key characteristic parameters of the culvert. The collected raw data is then processed and encoded inside the deformation sensor to form a structured Bluetooth data packet. This data packet follows the Bluetooth communication protocol to ensure the integrity and correctness of the data during transmission. After the encoding is completed, the deformation sensor uses Bluetooth wireless technology to send the data packet to the processor. After the Bluetooth receiving module of the processor receives the data packet, it performs a decoding operation to parse out the culvert deformation data in the data packet. The decoding process includes verifying the integrity of the data packet, performing error detection and verification, and extracting useful data content.
[0032] During the entire process, communication stability and data accuracy are crucial, and software settings are required to optimize the data transmission rate and retransmission mechanism to ensure the reliability of data transmission.
[0033] In an optional embodiment, the processor obtains the current deformation data of the culvert to be tested under the real-time pressure detected by the deformation sensor, specifically: The deformation sensor establishes a communication connection with the processor through the RS-232 serial interface. In this process, it is first necessary to ensure that the deformation sensor and the RS-232 port of the processor are properly connected, and configure the corresponding communication parameters, such as baud rate, data bits, stop bits, and parity bits, to match the communication settings of both parties. Once the communication settings are completed, the processor sends a data acquisition signal instruction to the deformation sensor through the RS-232 interface. This instruction is a specific data sequence used to tell the deformation sensor to start the data acquisition process. After receiving the instruction, the deformation sensor activates the internal data acquisition mechanism and starts measuring the deformation data of the culvert. The collected data is then encoded inside the deformation sensor and constructed into a data packet that conforms to the communication protocol. This data packet includes a start bit, data content, check bit, and stop bit to ensure the integrity and correctness of the data. After the encoding is completed, the deformation sensor sends the data packet back to the processor through the RS-232 bus. After the RS-232 interface of the processor receives the data packet from the deformation sensor, it performs decoding processing. The decoding process involves checking the structure of the data packet, performing error detection and verification, and then extracting the actual deformation data.
[0034] The entire communication process requires precise time synchronization and data synchronization to ensure data accuracy and reliability; in addition, in order to improve the stability and anti-interference ability of communication, some additional measures will be taken, such as using shielded cables, signal boosters and error retransmission mechanisms.
[0035] In step S12, the current deformation data is preprocessed to obtain a current deformation vector.
[0036] It should be noted that the deformation sensor described in this embodiment includes two strain gauge sensors, which are respectively arranged at the top and bottom of the culvert to measure the strain value of the top and the annular stress strain value of the bottom; wherein the current deformation data is the absolute value average of the strain value of the top and the strain value of the bottom.
[0037] In an optional embodiment, the preprocessing the current deformation data to obtain the current deformation vector includes: A plane rectangular coordinate system is constructed, with the current deformation data as the vertical axis and time as the horizontal axis, to plot a strain curve of the current deformation data changing with time, wherein the 0 point of the horizontal axis is the moment when the periodic changing pressure begins to be applied; a number of feature points are evenly taken out within the 0-T time of the plane rectangular coordinate system, and the current deformation data corresponding to each feature point is obtained based on the strain curve; and a current deformation vector is constructed based on all the current deformation data.
[0038] It should be noted that the selection of feature points is to uniformly select data points from the strain curve within a given time period (0-T), and these points are subsequently used to construct the current deformation vector. Feature points need to be evenly distributed, which means that the time interval between each feature point should be equal. For example, if the time period T is 10 seconds and it is decided to select 5 feature points, then these points should be obtained at 0 seconds, 2.5 seconds, 5 seconds, 7.5 seconds and 10 seconds respectively. The selected feature points should be able to represent the changing trend of the deformation data in the entire period. Uniformly distributed feature points can ensure the representativeness and consistency of deformation data, avoid data being too dense or sparse in certain time periods, and cause the model to inaccurately capture the changing trend of data during training. Uniformly selected feature points make each feature point equally representative of the deformation data in the entire time period, thereby better reflecting the deformation characteristics in the entire period; based on the strain curve, these evenly distributed feature points are accurate records of the current deformation data, which are obtained by checking the curve position at a specific time point. The obtained deformation data value will be directly used to construct the deformation vector, which is a numerical array, each value represents the deformation data value of a feature point on the strain curve. Uniform feature point selection ensures the accuracy of the deformation vector and the overall performance of the model.
[0039] Furthermore, determining the number of feature points to be selected usually depends on the complexity of the data, the purpose of the analysis, and the desired accuracy. First, the changing characteristics of the strain curve need to be considered. If the curve changes relatively slowly or has obvious periodicity, too many feature points may not be needed to capture the trend; if the curve fluctuates greatly or contains multiple change modes, more feature points are needed to ensure that all important changes are captured. Secondly, the purpose of the analysis is also very critical. If it is for high-precision modeling or prediction, more data points may be needed to build a detailed model; if it is just to observe the general trend, fewer points may be sufficient. Usually, you can start with a trial and error method, select a baseline number of points, conduct a preliminary analysis, and then adjust the number of points based on the results to find the best balance. Ultimately, the number of feature points should be able to reflect all key information while avoiding overfitting or waste of computing resources.
[0040] In step S13, the current deformation vector is input into a pre-trained deep learning model to obtain a reconstruction error.
[0041] In an optional embodiment, the deep learning model uses an autoencoder model. The autoencoder is an unsupervised learning neural network used to learn effective encoding of data. In terms of network structure, the autoencoder consists of an encoder and a decoder. The encoder maps the input data to a low-dimensional implicit representation (encoding), and gradually compresses information through multiple hidden layers to achieve the effect of data dimensionality reduction. The decoder reconstructs the input data from this implicit representation and restores the low-dimensional representation to the original data space through reverse operations. The training goal of the autoencoder is to make the reconstructed data as close as possible to the original input data, so its loss function is usually the reconstruction error between the input data and the reconstructed data. The network parameters are adjusted through the back propagation algorithm so that the combination of the encoder and the decoder can effectively capture the main features and structure of the data. The autoencoder can also be extended in different ways, such as sparse autoencoders (by adding sparsity constraints to improve feature extraction capabilities), variational autoencoders (VAE, by introducing a probability model to generate new data samples), etc. These variants further enhance the application scope and effect of autoencoders in data representation learning and generation tasks. The autoencoder is trained by input data, with the goal of minimizing the difference between input and output, thereby learning a low-dimensional representation of the data. The goal of the autoencoder is to make the reconstructed output as close to the original input as possible, thereby ensuring that the encoding process captures the main features of the data. The feature extraction function of the autoencoder model can automatically learn high-level features in the data, which are extremely important for subsequent tasks such as classification and prediction.
[0042] Specifically, refer to Figure 2 , the training process of the autoencoder model includes: S21, obtaining a first sample deformation data set and a second sample deformation data set; S22, preprocessing the first sample deformation data set and the second sample deformation data set to obtain a first sample vector set and a deformed second sample vector set; S23, inputting the first sample vector set into an initial deep learning model to train the initial deep learning model; S24, input the second sample vector set into the initial deep learning model, evaluate the model, and obtain a comprehensive evaluation index. When the comprehensive evaluation index meets the preset evaluation index, a trained deep learning model is obtained.
[0043] In step S21, a first sample deformation data set and a second sample deformation data set are obtained.
[0044] It should be noted that the first sample deformation data set is measured in advance from a batch of normally working culverts; the second sample deformation data set is measured in advance from a batch of culverts including normally working culverts and abnormally working culverts, wherein the actual working condition of each culvert has been obtained and marked in advance.
[0045] In step S24, the second sample vector set is input into the initial deep learning model, the model is evaluated to obtain a comprehensive evaluation index, and when the comprehensive evaluation index meets the preset evaluation index, the trained deep learning model is obtained.
[0046] In an optional embodiment, the step of inputting the second sample vector set into an initial deep learning model, evaluating the model, obtaining a comprehensive evaluation index, and obtaining a trained deep learning model when the comprehensive evaluation index satisfies a preset evaluation index includes: The second sample vector set is input into the initial deep learning model for detection, and the reconstruction error corresponding to each culvert corresponding to the second sample vector set is obtained; the reconstruction error is compared with the error threshold to determine whether the culvert is working properly; the judgment result is compared with the actual working condition of the culvert, and the number TP of normal culverts judged as normal culverts, the number FN of normal culverts judged as abnormal culverts, and the number FP of abnormal culverts judged as normal culverts are calculated; the comprehensive evaluation index is compared with the preset standard evaluation index. If the comprehensive evaluation index is greater than the standard evaluation index, it means that the model has completed the feature learning of the working normal culvert, and the trained deep learning model is obtained. If the comprehensive evaluation index is less than the threshold, it means that the model has not completed the feature learning of the working normal culvert, and it is necessary to re-collect the sample vector set and input it into the model for training.
[0047] Furthermore, the process of calculating the comprehensive evaluation index includes:
[0048]
[0049]
[0050] Among them, F1 is a comprehensive evaluation indicator, the range of F1 is 0 to 1, 1 represents perfect performance, 0 represents the worst performance, P is the precision rate, and R is the recall rate.
[0051] In step S13, the current deformation vector is input into a pre-trained deep learning model to obtain a reconstruction error.
[0052] It should also be noted that the reconstruction error is an indicator used to measure the accuracy of the model's reconstruction data. The reconstruction error essentially measures the difference between the input data and the output data reconstructed by the model.
[0053] Furthermore, the calculation method of the reconstruction error depends on the type of data and the application scenario. Where is the original data point, corresponding to the characteristic point in the pressure deformation curve formed by the current deformation vector input, is the reconstructed data point, which is the feature point extracted by the model based on the features of the first sample vector set, and n is the total number of selected data points, including: Mean Squared Error (MSE):
[0054] This method is suitable for continuous numerical data and can directly measure the squared differences between data points.
[0055] Cross-Entropy Loss:
[0056] This method is suitable for binary classification problems, where and Usually represents probability (for example, the value of an image pixel is between 0 and 1), the cross entropy loss measures the difference between two probability distributions and is often used to process data with probabilistic nature.
[0057] Mean Absolute Error (MAE):
[0058] Applicable to scenarios where you need to measure the absolute difference between the actual value and the predicted value.
[0059] Cosine Similarity Loss:
[0060] It is usually used to compare the directional similarity of two vectors rather than their size, and is particularly suitable for similarity comparison of text data or high-dimensional data.
[0061] Reconstruction error is a key indicator to measure the accuracy of data reconstruction by deep learning models. It is calculated by comparing the difference between input data and model output data. In the field of culvert operation monitoring, this indicator is particularly important because it is directly related to the integrity and functional status of the culvert structure and can quantify the difference between input and output data. In the present invention, the input data is the deformation vector of the culvert to be tested.
[0062] In step S14, the reconstruction error is compared with a preset error threshold. When the reconstruction error is smaller than the error threshold, it is determined that the culvert is working normally; when the reconstruction error is larger than the error threshold, it is determined that the culvert is working abnormally.
[0063] In an optional embodiment, the processor compares the reconstruction error with a preset error threshold to evaluate the current state of the culvert. When the reconstruction error is less than the error threshold, it means that the deviation between the pressure deformation curve of the current culvert and the pressure deformation curve of the normal culvert is within an acceptable range, which indicates that the structure and performance of the current culvert meet the design and operation requirements. In this case, the culvert is considered to be normal, with no obvious signs of damage or deterioration, and it is safe and reliable to continue using it. On the contrary, when the reconstruction error is greater than the error threshold, it means that the deviation between the pressure deformation curve of the current culvert and the pressure deformation curve of the normal culvert exceeds the preset range. In this case, the culvert may have structural problems or be affected by external factors, resulting in significant changes in its deformation characteristics. Possible reasons include excessive external loads, material aging, geological activities, etc. The reconstruction error exceeding the error threshold indicates that the performance of the culvert no longer meets the design and operation requirements, and further inspection and maintenance are required to avoid potential safety hazards and operational failures.
[0064] Through this method, the status of the culvert can be monitored in real time, potential problems can be discovered in time and corresponding measures can be taken to ensure the safe operation of the culvert. The setting of the error threshold needs to be comprehensively considered based on the design specifications, historical data and actual use environment of the culvert to ensure the sensitivity and accuracy of the detection. At the same time, this monitoring method based on reconstruction error is efficient and reliable, and can provide continuous status assessment during the operation of the culvert, which helps to extend the service life of the culvert and improve the reliability of the overall system.
[0065] It should be noted that the preset error threshold is a specific value set according to the design specifications and safe operation requirements of the culvert system, which is used to define the normal working state and potential abnormal state of the culvert. This threshold is usually derived based on historical data analysis, system tolerance and engineering experience to ensure that the system operates without affecting its normal function and safety. If the reconstruction error exceeds this threshold, it may indicate that the culvert has structural damage, functional failure or other problems that affect system performance, and further inspection or maintenance is required. On the contrary, if the reconstruction error is less than this threshold, it can be considered that the culvert is in normal working condition and no special intervention is required. The preset error threshold is a key parameter that determines the state of the culvert when an alarm is issued to the user. In this embodiment, the preset error threshold is 1.0, which is a result verified by experiments and ensures that the culvert works under certain safety conditions. However, the error threshold is not fixed and may need to be adjusted accordingly according to the characteristics of the specific application scenario and the needs of the actual application. For example, the error threshold can also be set to 0.8, 1.2, etc., which is not limited by the present invention.
[0066] Furthermore, the method for obtaining a preset error threshold includes: obtaining the preset error threshold by importing a pre-configured external file; storing the preset error threshold as an external file and importing it when needed, thereby reducing memory and CPU consumption; or, the preset error threshold can be stored as an internal file, and the preset error threshold can be obtained by directly reading the pre-configured internal file; or, the preset error threshold can be obtained by instant call, and the preset error threshold can be obtained by instant adjustment of a parameter according to time.
[0067] The working process of the present invention is described below by taking a more specific scenario as an example.
[0068] In this example, the culvert is a cement culvert with a diameter of 1.5 meters and a length of two meters. Two strain sensors are installed in the center of the top and bottom of the inner wall of the culvert. The current deformation data is the average of the absolute values of the two strain sensors. The processor is an industrial computer, and the deep learning model used is an autoencoder.
[0069] The working process is as follows: Step 1: Selecting a first sample culvert and a second sample culvert from the culverts; Step 2: The industrial computer obtains a first sample deformation data set and a second sample deformation data set from the strain sensor; Step 3: The industrial computer preprocesses the first sample deformation data set and the second sample deformation data set to obtain a first sample vector set and a deformation second sample vector set; Step 4: The industrial computer inputs the first sample vector set into the autoencoder model to train the autoencoder model; Step 5: Input the second sample vector set into the autoencoder model and evaluate the model; Step 6: The industrial computer obtains the deformation vector of the culvert to be tested from the strain sensor; Step 7: The industrial computer inputs the deformation vector of the culvert to be tested into the trained autoencoder model, compares the reconstruction error output by the model with the preset threshold, and determines the normality of the culvert.
[0070] The culvert operation monitoring scheme of the present invention has the following advantages: by selecting the first and second sample culverts, the diversity and representativeness of the data are ensured, which is helpful for the comprehensive training of the model; by using the industrial computer to obtain the data of the strain sensor, high-precision and real-time deformation data collection can be achieved; the data is preprocessed and converted into a vector set, which enhances the operability of the data and the processing efficiency of the model; the use of the autoencoder model for training and evaluation helps to extract features in the deformation data and perform effective dimensionality reduction, improve the generalization ability of the model and the sensitivity to anomalies; by comparing the reconstruction error and the preset threshold, the normality of the culvert can be judged quickly and accurately, reducing manual intervention and improving the automation level and reliability of monitoring; in addition, the scheme can adapt to complex and changeable environmental conditions and ensure the long-term stability and continuity of culvert deformation monitoring.
[0071] Reference Figure 3 The second embodiment of the present invention provides a culvert operation monitoring device, comprising: A data acquisition module is used to obtain the current deformation data of the culvert to be tested under the real-time pressure detected by the deformation sensor; A data preprocessing module preprocesses the current deformation data to obtain a current deformation vector; An error calculation module inputs the current deformation vector into a pre-trained deep learning model to obtain a reconstruction error; A condition identification module compares the reconstruction error with a preset error threshold, and when the reconstruction error is less than the error threshold, it is determined that the culvert is working normally; when the reconstruction error is greater than the error threshold, it is determined that the culvert is working abnormally; A model training module obtains a first sample deformation data set and a second sample deformation data set; preprocesses the first sample deformation data set and the second sample deformation data set to obtain a first sample vector set and a deformed second sample vector set; inputs the first sample vector set into an initial deep learning model to train the initial deep learning model; inputs the second sample vector set into the initial deep learning model to evaluate the model to obtain a comprehensive evaluation index, and when the comprehensive evaluation index meets the preset evaluation index, a trained deep learning model is obtained.
[0072] It should be noted that the culvert operation monitoring device provided in the embodiment of the present invention is used to execute all the process steps of the culvert operation monitoring method of the above embodiment, and the working principles and beneficial effects of the two correspond one to one, so they will not be described in detail.
[0073] The embodiment of the present invention further provides an electronic device. The electronic device includes: a processor, a memory, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps in the above-mentioned culvert operation monitoring method embodiments are implemented, for example Figure 1 Alternatively, the processor implements the functions of each module / unit in the above-mentioned device embodiments when executing the computer program.
[0074] Exemplarily, the computer program may be divided into one or more modules / units, which are stored in the memory and executed by the processor to implement the present invention. The one or more modules / units may be a series of computer program instruction segments capable of implementing specific functions, which are used to describe the execution process of the computer program in the electronic device.
[0075] The electronic device may be a computing device such as a desktop computer, a notebook, a PDA, and a smart tablet. The electronic device may include, but is not limited to, a processor and a memory. Those skilled in the art will appreciate that the above components are merely examples of electronic devices and do not constitute a limitation on the electronic device. The electronic device may include more or fewer components than the above components, or may combine certain components, or different components. For example, the electronic device may also include input and output devices, network access devices, buses, etc.
[0076] The processor may be a central processing unit (CPU), other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor, etc. The processor is the control center of the electronic device, and uses various interfaces and lines to connect various parts of the entire electronic device.
[0077] The memory can be used to store the computer program and / or module, and the processor realizes various functions of the electronic device by running or executing the computer program and / or module stored in the memory, and calling the data stored in the memory. The memory can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, an application required for at least one function (such as a sound playback function, an image playback function, etc.), etc.; the data storage area can store data created according to the use of the mobile phone (such as audio data, a phone book, etc.), etc. In addition, the memory can include a high-speed random access memory, and can also include a non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), at least one disk storage device, a flash memory device, or other volatile solid-state storage devices.
[0078] Wherein, if the module / unit integrated in the electronic device is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on such an understanding, the present invention implements all or part of the processes in the above-mentioned embodiment method, and can also be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium, and the computer program can implement the steps of the above-mentioned various method embodiments when executed by the processor. Wherein, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording medium, U disk, mobile hard disk, disk, optical disk, computer memory, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), electric carrier signal, telecommunication signal and software distribution medium. It should be noted that the content contained in the computer-readable medium can be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media do not include electric carrier signals and telecommunication signals.
[0079] It should be noted that the device embodiments described above are merely schematic, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. In addition, in the accompanying drawings of the device embodiments provided by the present invention, the connection relationship between the modules indicates that there is a communication connection between them, which may be specifically implemented as one or more communication buses or signal lines. A person of ordinary skill in the art may understand and implement it without paying any creative effort.
[0080] The specific embodiments described above further describe the purpose, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. It is particularly pointed out that for those skilled in the art, any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention should be included in the scope of protection of the present invention.
Claims
1. A culvert operation monitoring method, characterized in that: Executed by the processor, including: Acquire the current deformation data of the culvert to be tested under the real-time pressure detected by the deformation sensor; Preprocessing the current deformation data to obtain a current deformation vector; Inputting the current deformation vector into a pre-trained deep learning model to obtain a reconstruction error; Comparing the reconstruction error with a preset error threshold, when the reconstruction error is less than the error threshold, judging that the culvert is working normally; when the reconstruction error is greater than the error threshold, judging that the culvert is working abnormally; The training process of the deep learning model includes: Acquire a first sample deformation data set and a second sample deformation data set; Preprocessing the first sample deformation data set and the second sample deformation data set to obtain a first sample vector set and a deformed second sample vector set; Inputting the first sample vector set into an initial deep learning model to train the initial deep learning model; The second sample vector set is input into the initial deep learning model, the model is evaluated to obtain a comprehensive evaluation index, and when the comprehensive evaluation index meets the preset evaluation index, a trained deep learning model is obtained.
2. The culvert operation monitoring method according to claim 1, characterized in that: The step of obtaining the current deformation data of the culvert to be tested under the real-time pressure detected by the deformation sensor includes: Receiving the coded real-time deformation quantity of the culvert to be tested sent by the deformation sensor, decoding the coded real-time deformation quantity of the culvert to be tested, and obtaining current deformation data corresponding to the culvert to be tested; The deformation sensor detects the real-time amount of the culvert deformation, encodes the real-time amount of the culvert deformation, and transmits the encoded real-time amount of the culvert deformation to the processor.
3. The culvert operation monitoring method according to claim 2, characterized in that: The real-time pressure is the pressure applied to the top of the culvert pipe, which changes in a sinusoidal or cosine manner according to time, with a period of T.
4. The culvert operation monitoring method according to claim 2, characterized in that: The deformation sensor comprises two strain gauge sensors, which are respectively arranged at the top and bottom of the culvert and are used to measure the strain value of the top and the hoop stress strain value of the bottom. Among them, the current deformation data is the absolute value average of the strain value at the top of the tube and the strain value at the bottom of the tube.
5. The culvert operation monitoring method according to claim 2, characterized in that: The preprocessing of the current deformation data to obtain a current deformation vector includes: Construct a plane rectangular coordinate system, with the current deformation data as the vertical axis and time as the horizontal axis, and draw a strain curve of the current deformation data changing with time, where the 0 point of the horizontal axis is the moment when the periodic pressure starts to be applied; Uniformly extracting a number of characteristic points within the 0-T time of the plane rectangular coordinate system, and obtaining current deformation data corresponding to each of the characteristic points based on the strain curve; The current deformation vector is constructed based on all the current deformation data.
6. The culvert operation monitoring method according to claim 2, characterized in that: The first sample deformation data set and the second sample deformation data set are specifically: The first sample deformation data set is pre-measured from a batch of normally working culverts; The second sample deformation data set is pre-measured from a batch of culverts including normally working culverts and abnormally working culverts, wherein the actual working condition of each culvert has been pre-obtained and marked.
7. The culvert operation monitoring method according to claim 2, characterized in that: The step of inputting the second sample vector set into the initial deep learning model, evaluating the model, obtaining a comprehensive evaluation index, and obtaining a trained deep learning model when the comprehensive evaluation index meets a preset evaluation index comprises: Inputting the second sample vector set into the initial deep learning model for detection, and obtaining a reconstruction error corresponding to each culvert corresponding to the second sample vector set; Comparing the reconstruction error with an error threshold to determine whether the culvert is working properly; Compare the judgment result with the actual working condition of the culvert, calculate the number of normal culverts judged as normal culverts TP, the number of normal culverts judged as abnormal culverts FN, and the number of abnormal culverts judged as normal culverts FP; Compare the comprehensive evaluation index with the preset standard evaluation index. If the comprehensive evaluation index is greater than the standard evaluation index, it means that the model has completed the feature learning of the normal working culvert, and the trained deep learning model is obtained. If the comprehensive evaluation index is less than the threshold, it means that the model has not completed the feature learning of the normal working culvert, and it is necessary to collect the sample vector set again and input it into the model for training; The calculation process of comprehensive evaluation index includes: Among them, F1 is a comprehensive evaluation indicator, the range of F1 is 0 to 1, 1 represents perfect performance, 0 represents the worst performance, P is the precision rate, and R is the recall rate.
8. A culvert operation monitoring device, characterized in that: include: A data acquisition module is used to obtain the current deformation data of the culvert to be tested under the real-time pressure detected by the deformation sensor; A data preprocessing module preprocesses the current deformation data to obtain a current deformation vector; An error calculation module inputs the current deformation vector into a pre-trained deep learning model to obtain a reconstruction error; A condition identification module compares the reconstruction error with a preset error threshold, and when the reconstruction error is less than the error threshold, it is determined that the culvert is working normally; when the reconstruction error is greater than the error threshold, it is determined that the culvert is working abnormally; A model training module obtains a first sample deformation data set and a second sample deformation data set; Preprocessing the first sample deformation data set and the second sample deformation data set to obtain a first sample vector set and a deformed second sample vector set; Inputting the first sample vector set into an initial deep learning model to train the initial deep learning model; The second sample vector set is input into the initial deep learning model, the model is evaluated to obtain a comprehensive evaluation index, and when the comprehensive evaluation index meets the preset evaluation index, a trained deep learning model is obtained.
9. An electronic device, characterized in that: The method comprises a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor implements the culvert operation monitoring method according to any one of claims 1 to 7 when executing the computer program.
10. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a stored computer program, wherein when the computer program is executed, the device where the computer-readable storage medium is located is controlled to execute the culvert operation monitoring method according to any one of claims 1 to 7.
Citation Information
Patent Citations
Low-temperature mechanical property testing method of superconducting conductor
CN110779818A
Water distribution system network physical attack detection method based on deep learning
CN113194098A
Intelligent monitoring method and system for compressed air pipeline
CN117494031A
Road and bridge settlement survey method and system
CN117723014A
Aviation flight data abnormity real-time detection method
CN117831351A
Cited By
Bamboo chopstick production line data real-time intelligent scheduling management system
CN119692736A
A real-time intelligent scheduling and management system for bamboo chopstick production line data
CN119692736B
