Shield machine operation and maintenance management system based on data analysis

Abstract

The present invention discloses a shield machine operation and maintenance management system based on data analysis, and relates to the technical field of shield equipment monitoring. The present invention comprises a data acquisition unit, a data transmission unit, a data processing and analysis unit, an anomaly detection unit, a verification unit and an early warning unit; the data acquisition unit is used to install a variety of sensors on the cutter part of the shield machine to collect cutting force, vibration, temperature, and displacement data to provide basic information for subsequent analysis; the present invention realizes real-time monitoring and anomaly detection of the state of the shield machine cutter by integrating a variety of sensors and a deep learning algorithm. Compared with the traditional manual inspection method, the system can provide more continuous and accurate monitoring, and reduce the errors and omissions caused by manual inspection; through automated data acquisition, transmission, processing and analysis, the efficiency and accuracy of shield machine operation and maintenance management are significantly improved, and the construction risks and economic losses caused by cutter anomalies are reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of shield equipment monitoring, and in particular to a shield machine operation and maintenance management system based on data analysis. Background Art

[0002] As the core equipment of tunnel construction, shield machines are widely used in many tunnel construction projects such as subways, railways, highways, and municipal pipelines. They can efficiently and safely carry out tunnel excavation operations under complex geological conditions, greatly improving construction efficiency and reducing construction risks. However, shield machines face many severe challenges during operation. Due to their harsh working environment and long-term exposure to high temperature, high pressure, high humidity, and strong vibration, the various parts of the equipment are prone to wear, fatigue, corrosion, and other problems, resulting in frequent failures. These failures will not only seriously affect the construction progress and cause delays in the construction period, but may also cause safety accidents, posing a threat to the lives of construction workers and causing huge economic losses.

[0003] During the construction of the shield machine, the cutter is a key component that directly contacts the rock and soil and performs cutting operations. Its working environment is extremely harsh, and it is subjected to huge impact force, friction and high temperature. Once the cutter is abnormal, such as wear, breakage, eccentric wear, etc., if it is not discovered and handled in time, it will lead to a significant decrease in the tunneling efficiency of the shield machine, and may even cause engineering accidents, resulting in huge economic losses and construction delays.

[0004] The traditional shield machine tool monitoring method mainly relies on manual regular inspection. This method has many defects: First, manual inspection can only be carried out when the shield machine is shut down, which not only affects the construction progress, but also during the interval between two inspections, the tool may have been seriously damaged without being detected in time. Secondly, the accuracy and reliability of manual inspections depend to a large extent on the experience and skill level of the inspectors. The judgments of different personnel may vary, and it is easy to miss or misjudge. Furthermore, manual inspections make it difficult to obtain real-time dynamic data of the tool during the excavation process, and it is impossible to continuously and accurately monitor and analyze the working status of the tool. It is difficult to predict the abnormal change trend of the tool in advance, and it is impossible to achieve preventive maintenance in the true sense.

[0005] With the rapid development of information technology, data analysis technology has shown great application potential in the field of industrial equipment operation and maintenance management. Through the real-time collection, transmission, storage and analysis of equipment operation data, accurate monitoring of equipment status, fault diagnosis and preventive maintenance can be achieved.

[0006] However, in the field of shield machine operation and maintenance management, the application of data analysis technology is still in its infancy. At present, although some shield machines have been equipped with some sensors to collect data, the data utilization efficiency is low and there is a lack of effective data analysis methods and models.

[0007] Existing data analysis systems are often unable to conduct in-depth mining and comprehensive analysis of massive, multi-source and heterogeneous data, making it difficult to extract valuable information to support accurate operation and maintenance decisions.

[0008] Therefore, there is an urgent need for a shield machine operation and maintenance management system based on data analysis to give full play to the advantages of data analysis technology, improve the level and efficiency of shield machine operation and maintenance management, and ensure the safe and stable operation of the shield machine.

[0009] In order to solve the above defects, a technical solution is now provided. Summary of the invention

[0010] The purpose of the present invention is to solve the problems of inaccurate fault judgment, untimely maintenance and lack of early warning and predictive maintenance caused by the traditional operation and maintenance management method relying on manual experience and regular maintenance when the shield machine is operating in a complex and harsh environment, and to propose a shield machine operation and maintenance management system based on data analysis.

[0011] The purpose of the present invention can be achieved through the following technical solutions:

[0012] The shield machine operation and maintenance management system based on data analysis includes:

[0013] Data acquisition unit, which is used to install various sensors on the shield machine cutter to collect cutting force, vibration, temperature, and displacement data to provide basic information for subsequent analysis;

[0014] The data processing and analysis unit pre-processes the raw data to improve the quality, uses wavelet analysis and PCA algorithms to mine and extract features, and provides key data support for anomaly detection;

[0015] The abnormality detection unit uses a deep learning algorithm combining CNN and LSTM to build a model based on the extracted feature data to judge the abnormal state of the tool and determine its type;

[0016] The verification unit collects images through the camera, and after the camera self-check mechanism, locates abnormal areas and conducts quality inspection, it compares the abnormal detection results with the normal images to ensure accuracy;

[0017] The early warning unit issues early warnings in different ways depending on the severity and type of the abnormality.

[0018] Furthermore, the specific steps performed by the verification unit are as follows:

[0019] When the model determines that the tool is in an abnormal state and determines the type of abnormality, the abnormal situation and type are verified; the full view and working state of the tool are captured by the camera installed around the cutterhead of the shield machine. The camera selection must adapt to the harsh working environment of the shield machine and perform regular self-inspection through the self-inspection mechanism; the camera transmits image data through a wired network;

[0020] Then obtain the abnormal type and area determined in the data processing and analysis unit, locate the corresponding abnormal area in the image data taken by the camera, and if the abnormal area is given in the form of coordinates, directly select the area in the image; if it is given in the form of feature description, use the image processing algorithm to perform feature matching and area search to find the area that matches the description;

[0021] Mark the located abnormal area by using borders, symbols or texts of different colors to indicate the location and scope of the abnormal area;

[0022] According to the marked abnormal area, a local image containing the abnormal area is extracted from the original image, and the abnormal area image is cropped based on the boundary of the abnormal area through image cropping operation; the extracted abnormal area image is saved separately, and classified and stored according to the abnormal type and timestamp information to establish an image database;

[0023] Perform quality checks on the images of the screened abnormal areas to ensure that the images are clear and complete, and that there is no image damage or information loss due to problems during transmission or processing;

[0024] Then, a normal image database is established. When the shield machine is initially running or the cutter is in a brand new state, normal images of various areas of the cutter are collected through cameras installed around the cutter head. During the collection, multiple collections are ensured under different working conditions to cover various normal states. The collected normal images are preprocessed, including denoising, graying, cropping and normalization operations, to extract image features, and these features are stored in the database together with the corresponding cutter area position information.

[0025] For the screened abnormal area images, the same feature extraction algorithm as the normal image is used to extract its feature vector; the feature vector of the abnormal area image is matched with the feature vector of the corresponding tool area position in the normal image database, and the Euclidean distance is used to calculate the similarity measure between the two, through the formula: Get the Euclidean distance d, where and are the i-th eigenvalues ​​of the abnormal region image and the normal image respectively, n is the dimension of the eigenvector; and the Euclidean distance d is used as the similarity measure;

[0026] The calculated similarity is evaluated according to the set similarity threshold. When the similarity is higher than the threshold, it is considered that the image of the abnormal area is similar to the normal image and there is a misjudgment. When the similarity is lower than the threshold, it is verified that the tool does have an abnormality and the type of abnormality is consistent with the previous model judgment.

[0027] Furthermore, the self-checking mechanism process in the verification module is as follows:

[0028] By comprehensively evaluating the image data collected by the camera, it is determined whether the camera is in a normal state and comprehensively analyzed using image parameters. Image parameters include:

[0029] Image gradient: The Sobel operator is used to calculate the gradient amplitude of the image. The larger the gradient amplitude, the clearer the edges and details in the image. The formula is: , and are the horizontal and vertical gradients, respectively. is an image, Represents the convolution operation; calculates the average gradient amplitude of the entire image as a clarity indicator;

[0030] Feature point matching: Use the feature point detection algorithm to detect feature points in multiple consecutive frames of images, and find matching feature point pairs through the feature point matching algorithm; for the matching feature point pairs, calculate their displacement deviations between different frames; calculate the mean and standard deviation of the displacement deviations of all feature points. The mean reflects the overall translation of the image, and the standard deviation reflects the degree of image jitter;

[0031] Noise evaluation: Calculate the standard deviation of the grayscale value of the local area of ​​the image, then take the average value of the entire image as the noise estimation value, set the noise threshold, and score based on the comparison between the noise estimation value and the threshold;

[0032] Lighting uniformity assessment: Calculate the grayscale histogram of the image, count the number of pixels at different grayscale levels, and calculate the entropy value of the grayscale histogram. The larger the entropy value, the more uniform the grayscale distribution.

[0033] The image parameter value is compared with the corresponding preset threshold to determine whether it exceeds the range. If it exceeds the range, it is determined that the image data captured by the camera is abnormal. Otherwise, a comprehensive evaluation is performed through various image parameters. Specifically:

[0034] The image parameter scores are weighted and summed according to the preset weights to obtain the comprehensive score of the camera. A comprehensive score threshold is set. If the comprehensive score is higher than the threshold, the camera is judged to be in normal state; if it is lower than the threshold, it is considered that there is a problem with the camera and further inspection or repair is required. At the same time, according to the specific circumstances of each score, the type of problem with the camera is analyzed, including: clarity exceeding the range is judged as dirty or damaged lens, and stability exceeding the range is judged as loose installation of the camera.

[0035] Furthermore, the specific operation steps of performing quality inspection on the screened abnormal area images in the verification module are as follows:

[0036] A comprehensive evaluation is performed through different image quality factors, including:

[0037] Pixel integrity score: calculates the proportion of valid pixels in the image;

[0038] Edge integrity assessment: Use edge detection algorithm to detect image edges, and then calculate edge continuity. Specifically, divide the edge image into several small blocks, and count the connection of edge pixels in each small block. If the edge pixels in a small block form a continuous curve or line segment, the score of the small block is 1; otherwise, the score is 0. Calculate the average of all small block scores as the edge integrity score.

[0039] High-frequency energy ratio: Perform Fourier transform on the image to convert the image from the spatial domain to the frequency domain. In the frequency domain, the high-frequency part corresponds to the details and edge information of the image. Calculate the ratio of high-frequency energy to total energy H. The specific process is as follows: First calculate the total energy of the high-frequency area in the frequency domain image , and then calculate the total energy of the entire frequency domain image ,but ; The higher the high-frequency energy ratio H, the richer the detail information contained in the image;

[0040] Signal-to-Noise Ratio: Calculate the signal-to-noise ratio of the image;

[0041] Assign corresponding weights to each image quality factor, calculate the comprehensive quality score of the image by weighted summation method, and compare the comprehensive quality score with the preset quality score threshold. If the comprehensive quality score is greater than the quality score threshold, the image quality is judged to meet the requirements. Otherwise, the image is judged to have quality problems, and further analyze the causes, including checking the transmission line, camera equipment or image acquisition environment.

[0042] Furthermore, the execution process of the early warning unit is as follows:

[0043] When the verification unit determines that the tool is abnormal, the early warning unit is activated immediately; different early warning methods are used according to the severity and type of the abnormality;

[0044] For minor abnormalities, the operating room display screen will warn: a striking warning pop-up window will pop up on the display screen in the shield machine operating room. The pop-up window is displayed with a yellow background. The pop-up window content includes tool number, abnormal type, specific value of wear, current temperature rise value, and vibration amplitude deviation value detailed information, so that the operator can understand the abnormal situation of the tool; at the same time, the position of the tool where the abnormality occurs is marked with a dynamic flashing icon on the main interface of the display screen, and the abnormal prompt information is displayed next to the icon;

[0045] Parameter adjustment suggestion prompt: Based on the abnormal situation analysis results, specific excavation parameter adjustment suggestions are provided on the display screen. The adjustment suggestions are presented in the form of a combination of text descriptions and dynamic charts; the tool status data is automatically updated at preset time intervals;

[0046] For moderate abnormalities, the sound and light alarm in the operation room is activated: the sound and light alarm is triggered in the shield machine operation room, the alarm sound is an intermittent buzzer, and the light is a yellow flashing light; the sound and light alarm lasts for 10-15 seconds, and it is triggered again after an interval of 5 seconds; when the sound and light alarm is triggered, the display screen in the operation room switches to a full-screen display of the tool abnormality specific information page, which contains all the information for minor abnormalities, and also displays the vibration frequency deviation value, wear area parameter, and the duration of the current abnormal state;

[0047] SMS notification sending: SMS notifications are automatically sent to on-duty technicians and on-site foremen on the ground. The SMS content includes the shield machine number, tool abnormality location, abnormality type, quantitative indicators of the abnormality degree, and currently recommended measures; SMS sending adopts priority setting, which is higher than work notifications but lower than emergency notifications; every 10 minutes, an updated tool abnormality information SMS is automatically sent until the abnormal situation is resolved or upgraded to a severe abnormality;

[0048] For severe abnormalities, strong sound and light alarms: A strong sound and light alarm is activated in the shield machine operation room. The alarm sound is a continuous sharp high-pitched sound, and the light is a red flashing strong light. The sound and light alarm continues until the operator manually confirms that the alarm has been received or the abnormal situation has been handled; at the same time, all non-critical equipment in the operation room, including some auxiliary lighting and air conditioning, is automatically turned off to concentrate power supply to critical systems; a red warning sign and text prompts are displayed on the display screen, and the same content is repeatedly played in the form of voice broadcast;

[0049] Multi-channel emergency notification: Immediately send emergency SMS notifications to ground engineering management personnel, equipment maintenance team leaders and relevant experts. The SMS content describes the severe abnormality of the tool, including all relevant quantitative data and the consequences, and requires relevant personnel to quickly organize forces to formulate maintenance plans and response measures; at the same time, send a detailed abnormality report to the above-mentioned personnel via email, and the report provides historical data trend charts and image analysis results of tool abnormalities in the form of attachments; automatically dial the preset emergency contact number and notify relevant personnel to handle the abnormality in the form of voice broadcast;

[0050] Automatic shutdown preparation: Send instructions to the shield machine's control system to prepare for the automatic shutdown procedure; before shutdown, gradually reduce the shield machine's excavation speed to the lowest safe speed; at the same time, adjust the cutter head speed to the lowest stable speed; during the shutdown process, continuously monitor the tool status and shield machine parameters; once the shutdown is completed, lock some operating functions of the shield machine, including prohibiting the restart of the excavation program and limiting the rotation of the cutter head to prevent misoperation.

[0051] Furthermore, the severity assessment process of tool abnormality in the early warning unit is as follows:

[0052] Obtain the parameters preprocessed by the data processing and analysis unit and perform preliminary analysis, including:

[0053] Wear index:

[0054] Tool residual thickness measurement: Regularly measure the tool's residual thickness. Suppose the tool's initial thickness is G0, and the currently measured residual thickness is G1, then the wear amount is G=G0-G1.

[0055] Wear area calculation: For tools with irregular wear shapes, the area of ​​the wear area is calculated through image analysis technology. The tool surface is photographed to obtain a high-resolution image. The wear area boundary is identified using an image processing algorithm, and then the wear area YG is calculated.

[0056] After normalizing the wear amount G and the wear area YG, the wear amount G and the wear area YG are used as the two right-angled sides of the right triangle, and the remaining side is connected to establish a right triangle. The area of ​​the right triangle is calculated and recorded as the wear evaluation value.

[0057] Set different wear evaluation value intervals according to tool materials and design requirements; determine the severity of wear based on the wear evaluation value intervals;

[0058] Vibration characteristic parameters:

[0059] Vibration amplitude analysis: The amplitude of the vibration signal is calculated through the tool vibration data, and the average or peak value of the vibration amplitude in the preset time period is used to measure it. The deviation from the average or peak value of the normal vibration amplitude is used to calculate the vibration deviation value;

[0060] Vibration frequency change: Perform spectrum analysis on the vibration signal to obtain the vibration frequency components, and calculate the vibration frequency deviation value through the vibration frequency components, which is recorded as the frequency deviation value;

[0061] After the obtained vibration deviation value and frequency deviation value are normalized, the bottom circle is drawn with the frequency deviation value as the bottom circle radius, and a cone model is established with the vibration deviation value as the height, and the volume of the cone model is calculated and recorded as the vibration balance value; a number of vibration balance value intervals are preset according to the vibration balance value, and the preset vibration balance value intervals correspond to different vibration abnormality levels, and the vibration abnormality level is judged by the vibration balance value;

[0062] Temperature change degree:

[0063] Tool temperature rise value: The temperature of the tool during operation is measured by a temperature sensor installed near the tool, while the ambient temperature is recorded to calculate the temperature rise value;

[0064] Temperature change rate: calculate the rate of change of tool temperature per unit time;

[0065] The obtained temperature rise value and temperature change rate are calibrated as ws and wb respectively, and after normalization, they are entered into the following formula: To get the temperature estimate ZO, where is a correction coefficient; the obtained temperature value ZO is used as a measure of the degree of temperature change; the obtained temperature value ZO is compared with a plurality of preset temperature value intervals, and different temperature anomaly levels are set corresponding to the plurality of temperature value intervals, and the temperature anomaly level is determined by comparison;

[0066] According to the importance of each indicator on the degree of tool abnormality, weights are assigned to the wear evaluation value, vibration balance value and temperature evaluation value. For each indicator, a score is given based on the comparison between its quantitative value and the set threshold. The comprehensive abnormality degree score is then calculated by the weighted summation method. The abnormality level is divided into slight abnormality, moderate abnormality and severe abnormality according to the comprehensive abnormality degree score.

[0067] Furthermore, the specific operation steps of the data acquisition unit are as follows:

[0068] Install a variety of sensors on the cutter of the shield machine, including pressure sensors, vibration sensors, temperature sensors and cutter displacement sensors;

[0069] The pressure sensor is used to detect the size and change of the cutting force that the tool bears during the process of cutting rock and soil;

[0070] The vibration sensor can capture the vibration frequency and amplitude information of the tool caused by wear and collision;

[0071] The temperature sensor monitors the temperature changes of the tool when it is working;

[0072] The tool displacement sensor can measure the displacement of the tool in different directions to determine whether there is any abnormality;

[0073] The sensor transmits the data collected in real time to the data transmission unit.

[0074] Furthermore, the specific operation steps of the data transmission unit are as follows:

[0075] A hybrid wired and wireless transmission network is used to ensure that data can be transmitted from the data acquisition unit to the data processing and analysis unit. Inside the shield machine, a wired network is used to connect sensors and data transmission nodes to ensure the stability and anti-interference of data transmission.

[0076] Wireless network is used for data transmission between the shield machine and the ground monitoring center to achieve remote and rapid data transmission.

[0077] Furthermore, the specific operation steps of the data processing and analysis unit are as follows:

[0078] After receiving the raw data from the data transmission unit, data preprocessing is performed, including data cleaning, denoising, and data format conversion operations to remove outliers and interference factors in the data;

[0079] Data analysis algorithms are used to conduct in-depth data mining and feature extraction. Specifically, wavelet analysis algorithms are used to perform time-frequency analysis on vibration data and extract vibration characteristics of different frequency bands. Principal component analysis algorithms are used to reduce the dimensionality of multi-dimensional sensor data and extract key feature vectors to reduce the complexity of data processing and highlight the characteristic information of the tool working status.

[0080] Furthermore, the specific operation steps of the abnormality detection unit are as follows:

[0081] Based on the feature data extracted by the data processing and analysis unit, the anomaly detection unit builds an intelligent anomaly detection model, which specifically adopts a deep learning algorithm and uses a combination of a convolutional neural network and a long short-term memory network. The convolutional neural network is used to extract the spatial features of the data and identify the local feature patterns in the sensor data. The long short-term memory network can learn the change law of the tool working state over time in processing time series data, thereby capturing the development trend of the tool abnormal state.

[0082] By inputting data on the normal and abnormal working conditions of known tools into the model for training, the model can automatically learn the characteristic patterns of the tools under different working conditions and establish corresponding discrimination criteria; when receiving new tool data, the model determines whether the tool is in an abnormal state and determines the type of abnormality, including wear, fracture, and eccentric wear.

[0083] Compared with the prior art, the present invention has the following beneficial effects:

[0084] The present invention realizes real-time monitoring and abnormality detection of shield machine cutter status by integrating multiple sensors and deep learning algorithms. Compared with the traditional manual inspection method, the system can provide more continuous and accurate monitoring, reducing errors and omissions caused by manual inspection; through automated data collection, transmission, processing and analysis, the efficiency and accuracy of shield machine operation and maintenance management are significantly improved, and the construction risks and economic losses caused by cutter abnormalities are reduced;

[0085] The present invention can predict the abnormal change trend of the tool in advance through in-depth analysis of the tool working state, thereby realizing preventive maintenance; the model constructed by combining CNN and LSTM algorithms can effectively identify the abnormal state and type of the tool, and the verification unit further ensures the accuracy of the detection result through image comparison. The early warning unit issues early warnings in different ways according to the severity and type of the abnormality, providing timely feedback to the operator, so that the maintenance work can be more timely and targeted, thereby reducing the possibility of failure.

[0086] The present invention provides a strong guarantee for the stable operation of the shield machine by comprehensively evaluating the wear amount, vibration characteristics and temperature change degree of the tool; the system can issue an early warning when the tool has a slight abnormality, and gradually upgrade the early warning measures as the degree of abnormality increases until safety measures such as automatic shutdown are taken, thereby ensuring the safety of the construction process; at the same time, the system's real-time monitoring and data analysis capabilities also improve the reliability of shield machine construction, reduce construction delays and engineering accidents caused by tool problems, and ensure the safety of construction personnel and equipment. BRIEF DESCRIPTION OF THE DRAWINGS

[0087] In order to facilitate understanding by those skilled in the art, the present invention is further described below in conjunction with the accompanying drawings;

[0088] Figure 1 This is the overall system block diagram of the present invention. DETAILED DESCRIPTION

[0089] The technical solution of the present invention will be clearly and completely described below in conjunction with the embodiments. 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.

[0090] It should be understood that the terms “include” and “comprising” used in the specification and claims of the present disclosure indicate the presence of described features, integers, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or collections thereof.

[0091] It should also be understood that the terms used in this disclosure are only for the purpose of describing specific embodiments and are not intended to limit the disclosure. As used in this disclosure and claims, the singular forms of "a", "an", and "the" are intended to include the plural forms unless the context clearly indicates otherwise. It should also be further understood that the term "and / or" used in this disclosure and claims refers to any combination of one or more of the associated listed items and all possible combinations, including these combinations.

[0092] like Figure 1 As shown, the shield machine operation and maintenance management system based on data analysis includes a data acquisition unit, a data transmission unit, a data processing and analysis unit, an anomaly detection unit, a verification unit and an early warning unit;

[0093] The data acquisition unit is used to install a variety of sensors on the shield machine cutter to collect data such as cutting force, vibration, temperature, displacement, etc., providing basic information for subsequent analysis;

[0094] A variety of sensors are installed on the tool part of the shield machine, including pressure sensors, vibration sensors, temperature sensors and tool displacement sensors; the pressure sensor is used to detect the size and change of the cutting force that the tool is subjected to during the process of cutting rock and soil; the vibration sensor can capture the vibration frequency and amplitude information of the tool caused by wear, collision, etc.; the temperature sensor monitors the temperature change of the tool during operation, and excessive temperature may indicate excessive friction or abnormal wear of the tool; the tool displacement sensor can accurately measure the displacement of the tool in different directions to determine whether there are abnormal conditions such as eccentric wear; these sensors transmit the data collected in real time to the data transmission unit.

[0095] The data transmission unit uses wired and wireless networks to ensure that data is transmitted stably and quickly from the acquisition unit to the processing and analysis unit, realizing remote data transmission and facilitating ground personnel to grasp the tool status in real time;

[0096] A hybrid wired and wireless transmission network is used to ensure that data can be transmitted from the data acquisition unit to the data processing and analysis unit in a timely and accurate manner. Inside the shield machine, a wired network is used to connect sensors and data transmission nodes to ensure the stability and anti-interference of data transmission. Between the shield machine and the ground monitoring center, a wireless network (such as 5G communication technology) is used for data transmission to achieve remote and rapid data transmission, so that ground technicians can understand the working status of the tool in real time.

[0097] The data processing and analysis unit pre-processes the raw data to improve its quality, and uses algorithms such as wavelet analysis and PCA to mine and extract features, providing key data support for anomaly detection;

[0098] After receiving the raw data from the data transmission unit, this unit first performs data preprocessing, including data cleaning, denoising, and data format conversion operations to remove outliers and interference factors in the data and improve the quality of the data; then, it uses data analysis algorithms to conduct in-depth data mining and feature extraction; specifically, it uses a wavelet analysis algorithm to perform time-frequency analysis on vibration data to extract vibration characteristics in different frequency bands; and uses the principal component analysis (PCA) algorithm to perform dimensionality reduction processing on multi-dimensional sensor data and extract key feature vectors to reduce the complexity of data processing and highlight the main feature information of the tool working status.

[0099] The abnormality detection unit uses a deep learning algorithm combining CNN and LSTM to build a model based on the extracted feature data to judge the abnormal state of the tool and determine its type;

[0100] Based on the feature data extracted by the data processing and analysis unit, the anomaly detection unit constructs an intelligent anomaly detection model, which specifically adopts a deep learning algorithm and uses a combination of a convolutional neural network (CNN) and a long short-term memory network (LSTM). CNN is good at extracting spatial features of data and can effectively identify local feature patterns in sensor data, such as vibration patterns of specific frequencies and the shape of pressure change curves. LSTM has advantages in processing time series data and can learn the changing laws of tool working conditions over time, thereby accurately capturing the development trend of tool abnormal conditions. By inputting a large amount of data on the normal and abnormal working conditions of known tools into the model for training, the model can automatically learn the feature patterns of tools under different working conditions and establish corresponding discrimination criteria. When receiving new tool data, the model can determine whether the tool is in an abnormal state and determine the type of abnormality, including wear, fracture, and eccentric wear.

[0101] The verification unit collects images through the camera, and after a series of processing (including self-checking, locating abnormal areas, quality inspection, etc.), compares them with normal images to verify the abnormal detection results and ensure accuracy;

[0102] When the model determines that the tool is in an abnormal state and determines the type of abnormality, the abnormal situation and type are verified. The specific steps are as follows:

[0103] The camera installed around the cutterhead of the shield machine captures the overall view and working status of the cutter. The camera should be waterproof, dustproof and shockproof to adapt to the harsh working environment of the shield machine. The self-check mechanism should be used to perform regular self-checks. Specifically: the image data collected by the camera is comprehensively evaluated to determine whether the camera status is normal. The image parameters are used for comprehensive evaluation. The image parameters include:

[0104] Image gradient: The Sobel operator is used to calculate the gradient amplitude of the image. The larger the gradient amplitude, the clearer the edges and details in the image. The formula is: , and are the horizontal and vertical gradients, respectively. is an image, Represents convolution operation; calculate the average gradient amplitude of the entire image as a clarity index. The higher the average gradient amplitude, the better the image clarity; feature point matching: use feature point detection algorithm (SIFT or SURF algorithm) to detect feature points in multiple consecutive frames, and find matching feature point pairs through feature point matching algorithm (distance-based matching method); for matching feature point pairs, calculate their displacement deviations between different frames. Count the mean and standard deviation of all feature point displacement deviations. The mean reflects the overall translation of the image, and the standard deviation reflects the degree of image jitter; noise evaluation: calculate the standard deviation of the grayscale value of the local area of ​​the image (such as 3×3 or 5×5 pixel blocks), then take the average value of the entire image as the noise estimation value, set the noise threshold, and score based on the comparison between the noise estimation value and the threshold; illumination uniformity evaluation: calculate the grayscale histogram of the image, count the number of pixels at different grayscale levels, and calculate the entropy value of the grayscale histogram. The larger the entropy value, the more uniform the grayscale distribution;

[0105] The above scores are compared with the corresponding preset thresholds to determine whether there is an over-range situation. If there is an over-range situation, it is determined that the image data captured by the camera is abnormal. Otherwise, a comprehensive evaluation is performed through various image parameters. Specifically:

[0106] The above scores are weighted and summed according to the preset weights to obtain the comprehensive score of the camera. A comprehensive score threshold is set. If the comprehensive score is higher than the threshold, the camera is judged to be in normal state; if it is lower than the threshold, it is considered that there is a problem with the camera and further inspection or repair is required. At the same time, according to the specific circumstances of each score, the types of problems that may exist in the camera are analyzed, including: clarity exceeding the range is judged as dirty or damaged lens, stability exceeding the range is judged as loose installation of the camera, so that targeted measures can be taken.

[0107] The camera can transmit image data stably and quickly through a wired network; then obtain the abnormal type and area determined in the data processing and analysis unit, locate the corresponding abnormal area in the image data taken by the camera, and if the abnormal area is given in the form of coordinates (such as the coordinates of the upper left corner and the lower right corner), directly select the area in the image; if it is given in the form of feature description (the approximate position of a specific shape or texture feature in the image), use the image processing algorithm to perform feature matching and area search to find the area that matches the description; mark the located abnormal area, use different colored borders, special symbols or annotated text to identify the position and range of the abnormal area for subsequent observation and analysis;

[0108] According to the marked abnormal area, extract the local image containing the abnormal area from the original image, and crop the abnormal area image based on the boundary of the abnormal area through image cropping operation, ensuring that the cropped image contains sufficient context information for further analysis of the abnormal situation; save the extracted abnormal area image separately, and classify and store it according to the abnormal type, timestamp and other information to establish an image database;

[0109] Perform quality checks on the images of the screened abnormal areas to ensure that the images are clear and complete, and that there is no image damage or information loss due to problems in the transmission or processing process; conduct a comprehensive assessment based on different image quality factors, including:

[0110] Pixel integrity score: calculate the proportion of valid pixels (non-zero pixel values) in the image; edge integrity assessment: use edge detection algorithms (such as Canny edge detection) to detect image edges, and then calculate the continuity of the edges. Specifically, divide the edge image into several small blocks (such as 5×5 pixel blocks), and count the connection of edge pixels in each small block; if the edge pixels in a small block form a continuous curve or line segment, the score of the small block is 1; otherwise, the score is 0; calculate the average of all small block scores as the edge integrity score; high-frequency energy proportion: perform Fourier transform on the image, convert the image from the spatial domain to the frequency domain. In the frequency domain, the high-frequency part corresponds to the details and edge information of the image, and calculate the proportion of high-frequency energy to the total energy H. The specific process is as follows: first calculate the total energy of the high-frequency area in the frequency domain image (such as a circular area with a radius greater than half the distance from the center of the image to the edge) , and then calculate the total energy of the entire frequency domain image ,but ; The higher the high-frequency energy ratio H, the richer the image details and the higher the image clarity; Signal-to-noise ratio: calculate the signal-to-noise ratio of the image. The higher the signal-to-noise ratio, the better the image quality and the smaller the impact of noise on the image;

[0111] Assign corresponding weights to the above-mentioned image quality factors, calculate the comprehensive quality score A of the image by the weighted sum method, and compare the comprehensive quality score A with the preset quality score threshold. If the comprehensive quality score A is greater than the quality score threshold, the image quality is judged to meet the requirements. Otherwise, the image is judged to have quality problems, and the causes are further analyzed, including checking the transmission line, camera equipment or image acquisition environment.

[0112] Then establish a normal image database. When the shield machine is initially running or the cutter is in a new state, the normal images of each area of ​​the cutter are collected through the cameras installed around the cutter head. During the collection, it should be ensured that multiple collections are carried out under different working conditions (different excavation speeds, geological conditions) to cover various possible normal states. The collected normal images are preprocessed, including denoising, graying, cropping and normalization operations, and the image features (edge ​​features, texture features, geometric features) are extracted, and these features are stored in the database together with the corresponding cutter area position information. For the screened abnormal area images, the same feature extraction algorithm as the normal image is used to extract its feature vector. The feature vector of the abnormal area image is matched with the feature vector of the corresponding cutter area position in the normal image database, and the similarity measurement between the two is calculated by the Euclidean distance, and the formula is: Get the Euclidean distance d, where and are the i-th eigenvalues ​​of the abnormal area image and the normal image respectively, n is the dimension of the eigenvector, the smaller the Euclidean distance d is, the higher the similarity is; according to the set similarity threshold, the calculated similarity is evaluated, when the similarity is higher than the threshold, it is considered that the abnormal area image is similar to the normal image and there is a misjudgment; when the similarity is lower than the threshold, it is verified that the tool does have an abnormality, and the abnormality type is consistent with the previous model judgment.

[0113] The warning unit issues warnings in different ways depending on the severity and type of the abnormality;

[0114] When the verification unit determines that the tool is abnormal, the early warning unit is activated immediately; different early warning methods are used according to the severity and type of the abnormality; the severity assessment process for the tool abnormality is as follows:

[0115] Obtain the parameters preprocessed by the data processing and analysis unit and perform preliminary analysis, including:

[0116] Wear index

[0117] Tool residual thickness measurement: Measure the tool's residual thickness regularly. Assume that the tool's initial thickness is G0, and the currently measured residual thickness is G1, then the wear amount is G=G0-G1; Wear area calculation: For tools with irregular wear shapes, calculate the area of ​​the wear area through image analysis technology, photograph the tool surface, obtain a high-resolution image, and use image processing algorithms (threshold segmentation, edge detection, etc.) to identify the wear area boundary, and then calculate the wear area area YG; After normalizing the wear amount G and the wear area area YG, use the wear amount G and the wear area area YG as the two right-angled sides of a right triangle, connect the remaining side length, establish a right triangle, calculate the area of ​​the right triangle, and record it as the wear evaluation value; set different wear evaluation value intervals according to the tool material and design requirements; judge the severity of wear according to the wear evaluation value interval;

[0118] Vibration characteristic parameters

[0119] Vibration amplitude analysis: Calculate the amplitude of the vibration signal through the tool vibration data, and use the average or peak value of the vibration amplitude in the preset time period to measure it. The vibration deviation value is obtained by calculating the deviation from the average or peak value of the normal vibration amplitude; Vibration frequency change: Perform spectrum analysis on the vibration signal to obtain the vibration frequency component, and calculate the vibration frequency deviation value through the vibration frequency component, which is recorded as the frequency deviation value; After normalizing the obtained vibration deviation value and frequency deviation value, draw the bottom circle with the frequency deviation value as the bottom circle radius, and establish a cone model with the vibration deviation value as the high value, calculate the volume of the cone model, and record it as the vibration balance value; preset several vibration balance value intervals according to the vibration balance value, and the preset several vibration balance value intervals correspond to different vibration abnormality levels, and the vibration abnormality level is judged by the vibration balance value;

[0120] Temperature change

[0121] Tool temperature rise value: The temperature of the tool during operation is measured by a temperature sensor installed near the tool, and the ambient temperature is recorded at the same time to calculate the temperature rise value; Temperature change rate: Calculate the rate of change of tool temperature per unit time; The obtained temperature rise value and temperature change rate are calibrated as ws and wb respectively, and after normalization, they are entered into the following formula: To get the temperature estimate ZO, where is a correction coefficient, and its value is set between 0.95 and 1.23; the obtained temperature value ZO is used as a measure of the degree of temperature change; the obtained temperature value ZO is compared with a number of preset temperature value intervals, and different temperature anomaly levels are set corresponding to the several temperature value intervals, and the temperature anomaly level is determined by comparison;

[0122] According to the importance of each indicator on the degree of tool abnormality, weights are assigned to the wear index, vibration characteristic parameters and temperature change degree. For each indicator, a score is given based on the comparison between its quantitative value and the set threshold. The comprehensive abnormality degree score is then calculated by the weighted summation method. The abnormality level is divided according to the comprehensive abnormality degree score; specifically, it is divided into slight abnormality, moderate abnormality and severe abnormality.

[0123] For minor abnormalities, the operating room display screen will warn: a striking warning pop-up window will pop up on the display screen in the shield machine operating room. The pop-up window is displayed with a yellow background (to distinguish warnings of different severity). The pop-up window content includes detailed information such as tool number, abnormality type (such as minor wear), specific value of wear, current temperature rise value, vibration amplitude deviation value, etc., so that the operator can intuitively understand the abnormality of the tool; at the same time, the position of the tool where the abnormality occurs is marked with a dynamic flashing icon (such as a yellow triangle) on the main interface of the display screen, and a brief abnormal prompt information is displayed next to the icon to ensure that the operator can notice the abnormality in time during the operation; parameter adjustment suggestion prompt: according to the analysis results of the abnormal situation, specific excavation parameter adjustment suggestions are provided on the display screen. The adjustment suggestions are presented in a combination of text descriptions and dynamic charts (such as a bar chart showing the comparison of parameters before and after adjustment) to facilitate the understanding and operation of the operator; every once in a while (such as 5 minutes), the system automatically updates the tool status data, and re-evaluates whether the parameter adjustment suggestions are reasonable based on the new data. If necessary, the adjustment parameters are further optimized to ensure that the tool is protected to the greatest extent without affecting the construction progress;

[0124] For moderate abnormalities, the sound and light alarm in the operation room is activated: the sound and light alarm is triggered in the shield machine operation room, the alarm sound is an intermittent buzzer (different from the continuous sharp alarm sound in an emergency), and the light is a yellow flashing light. The duration of the sound and light alarm is 10-15 seconds, and it is triggered again after an interval of 5 seconds to ensure that the operator can detect and pay attention to the abnormal situation in time, but will not interfere with normal work due to the long-term alarm sound; when the sound and light alarm is triggered, the display screen in the operation room switches to the full-screen display of the tool abnormality detailed information page. In addition to all the information for minor abnormalities, it also adds more parameters such as vibration frequency deviation value and wear area area, as well as the duration of the current abnormal state, so that the operator can have a more comprehensive understanding of the abnormality and development trend of the tool; SMS notification is sent: SMS notifications are automatically sent to the on-duty technicians and on-site foremen on the ground. The SMS content includes the shield machine number, the abnormal position of the tool, the abnormality type (moderate abnormality - specifically such as moderate wear, vibration abnormality, etc.), quantitative indicators of the degree of abnormality (such as specific values ​​such as wear, vibration balance value, temperature evaluation value, etc.) and the currently recommended measures (such as continued observation, preparation of spare tools, etc.). SMS messages are sent with priority settings to ensure that relevant personnel can receive notifications in a timely manner. The priority level is higher than general work notifications, but lower than emergency notifications. Updated tool abnormality information SMS messages are automatically sent every 10 minutes until the abnormal situation is resolved or upgraded to a severe abnormality.

[0125] For severe abnormalities, strong sound and light alarms: A strong sound and light alarm is activated in the shield machine operating room. The alarm sound is a continuous sharp high-pitched sound, and the light is a red flashing strong light. The sound and light alarm continues until the operator manually confirms that the alarm has been received or the abnormal situation has been handled; at the same time, all non-critical equipment in the operating room (such as some auxiliary lighting, air conditioning, etc.) is automatically shut down to concentrate power supply to key systems and attract the attention of operators; a huge red warning sign and a text prompt of "Serious abnormality of the cutter - stop the machine immediately" are displayed on the display screen, and the same content is repeated in the form of voice broadcast to ensure that the operator can understand the severity of the abnormality at the first time and take corresponding measures; multiple channels are used for emergency response. Emergency notification: Immediately send an emergency SMS notification to the engineering management personnel (project manager, technical person in charge, etc.) on the ground, the person in charge of the equipment maintenance team, and relevant experts. The SMS content describes in detail the severe abnormality of the tool, including all relevant quantitative data and possible serious consequences (such as possible shield machine failure, engineering accidents, etc.), and requires relevant personnel to quickly organize forces to formulate maintenance plans and countermeasures; at the same time, send a detailed abnormality report to the above-mentioned personnel via email, and the report provides the historical data trend chart of the tool abnormality (such as the wear amount over time curve, temperature change curve, etc.), image analysis results (such as wear area image) and preliminary maintenance suggestions automatically generated by the system in the form of attachments. In addition, the system will automatically call the preset emergency contact number (such as the project manager's mobile phone, the maintenance team's duty phone, etc.), and notify the relevant personnel in the form of voice broadcast to deal with the abnormal situation as soon as possible, to ensure the timeliness and accuracy of information transmission, and to avoid engineering accidents and delays caused by tool failure; automatic shutdown preparation: send instructions to the shield machine's control system to prepare to execute the automatic shutdown procedure. Before the shutdown, the tunneling speed of the shield machine is gradually reduced to the minimum safe speed; at the same time, the cutter head speed is adjusted to the minimum stable speed to smoothly transition to the shutdown state and reduce the impact on the shield machine and the surrounding soil; during the shutdown process, the tool status and parameters of other key systems of the shield machine are continuously monitored to ensure a safe and smooth shutdown process. Once the shutdown is completed, the system locks some operating functions of the shield machine (such as prohibiting the restart of the tunneling program, limiting the rotation of the cutter head, etc.) to prevent misoperation and wait for maintenance personnel to handle it.

[0126] The preferred embodiments of the present invention disclosed above are only used to help explain the present invention. The preferred embodiments do not describe all the details in detail, nor do they limit the invention to only specific implementation methods. Obviously, many modifications and changes can be made according to the content of this specification. This specification selects and specifically describes these embodiments in order to better explain the principles and practical applications of the present invention, so that those skilled in the art can understand and use the present invention well. The present invention is limited only by the claims and their full scope and equivalents.

Claims

1. The shield machine operation and maintenance management system based on data analysis is characterized by: include: Data acquisition unit, which is used to install various sensors on the shield machine cutter to collect cutting force, vibration, temperature, and displacement data to provide basic information for subsequent analysis; The data transmission unit uses wired and wireless networks to ensure data transmission from the data acquisition unit to the data processing and analysis unit, thus realizing remote data transmission; The data processing and analysis unit pre-processes the raw data to improve the quality, uses wavelet analysis and PCA algorithms to mine and extract features, and provides key data support for anomaly detection; The abnormality detection unit uses a deep learning algorithm combining CNN and LSTM to build a model based on the extracted feature data to judge the abnormal state of the tool and determine its type; The verification unit collects images through the camera, and after the camera self-check mechanism, locates the abnormal area and performs quality inspection, the abnormal detection results are compared with the normal image to ensure accuracy. The specific steps are as follows: When the model determines that the tool is in an abnormal state and determines the type of abnormality, the abnormal situation and type are verified; the full view and working state of the tool are captured by cameras installed around the cutterhead of the shield machine. The selection of cameras must adapt to the harsh working environment of the shield machine and self-check regularly through the self-check mechanism; The camera transmits image data through a wired network; Then obtain the abnormal type and area determined in the data processing and analysis unit, locate the corresponding abnormal area in the image data taken by the camera, and if the abnormal area is given in the form of coordinates, directly select the area in the image; if it is given in the form of feature description, use the image processing algorithm to perform feature matching and area search to find the area that matches the description; Mark the located abnormal area by using borders, symbols or texts of different colors to indicate the location and scope of the abnormal area; According to the marked abnormal area, a local image containing the abnormal area is extracted from the original image, and the abnormal area image is cropped based on the boundary of the abnormal area through image cropping operation; the extracted abnormal area image is saved separately, and classified and stored according to the abnormal type and timestamp information to establish an image database; Perform quality checks on the images of the screened abnormal areas to ensure that the images are clear and complete, and that there is no image damage or information loss due to problems during transmission or processing; Then, a normal image database is established. When the shield machine is initially running or the cutter is in a brand new state, normal images of various areas of the cutter are collected through cameras installed around the cutterhead. During the collection process, multiple collections are made under different working conditions to cover various normal states. Preprocess the normal image collected, including denoising, grayscale, cropping and normalization operations, extract the image features, and store these features together with the corresponding tool area position information in the database; For the screened abnormal area images, the same feature extraction algorithm as the normal image is used to extract its feature vector; The feature vector of the abnormal area image is matched with the feature vector of the corresponding tool area position in the normal image database, and the similarity measure between the two is calculated using the Euclidean distance, through the formula: Get the Euclidean distance d, where and are the i-th eigenvalues ​​of the abnormal region image and the normal image respectively, n is the dimension of the eigenvector; and the Euclidean distance d is used as the similarity measure; According to the set similarity threshold, the calculated similarity is evaluated. When the similarity is higher than the threshold, it is considered that the image of the abnormal area is similar to the normal image and there is a misjudgment; when the similarity is lower than the threshold, it is verified that the tool does have an abnormality. The self-checking mechanism process is as follows: By comprehensively evaluating the image data collected by the camera, it is determined whether the camera is in a normal state and comprehensively analyzed using image parameters. Image parameters include: Image gradient: The Sobel operator is used to calculate the gradient amplitude of the image. The larger the gradient amplitude, the clearer the edges and details in the image. The formula is: , and are the horizontal and vertical gradients, respectively. is an image, Represents the convolution operation; calculates the average gradient amplitude of the entire image as a clarity indicator; Feature point matching: Use the feature point detection algorithm to detect feature points in multiple consecutive frames of images, and find matching feature point pairs through the feature point matching algorithm; for the matching feature point pairs, calculate their displacement deviations between different frames; calculate the mean and standard deviation of the displacement deviations of all feature points. The mean reflects the overall translation of the image, and the standard deviation reflects the degree of image jitter; Noise evaluation: Calculate the standard deviation of the grayscale value of the local area of ​​the image, then take the average value of the entire image as the noise estimation value, set the noise threshold, and score based on the comparison between the noise estimation value and the threshold; Lighting uniformity assessment: Calculate the grayscale histogram of the image, count the number of pixels at different grayscale levels, and calculate the entropy value of the grayscale histogram. The larger the entropy value, the more uniform the grayscale distribution. The image parameter value is compared with the corresponding preset threshold to determine whether it exceeds the range. If it exceeds the range, it is determined that the image data captured by the camera is abnormal. Otherwise, a comprehensive evaluation is performed through various image parameters. Specifically: The image parameter scores are weighted and summed according to the preset weights to obtain the comprehensive score of the camera. A comprehensive score threshold is set. If the comprehensive score is higher than the threshold, the camera is considered to be in a normal state; if it is lower than the threshold, it is considered that there is a problem with the camera and further inspection or repair is required. At the same time, according to the specific situation of each score, the type of problem with the camera is analyzed, including: if the clarity exceeds the range, it is judged that the lens is dirty or damaged; if the stability exceeds the range, it is judged that the camera is loosely installed; The specific operation steps of performing quality inspection on the screened abnormal area images in the verification unit are as follows: A comprehensive evaluation is performed through different image quality factors, including: Pixel integrity score: calculates the proportion of valid pixels in the image; Edge integrity assessment: Use edge detection algorithm to detect image edges, and then calculate edge continuity. Specifically, divide the edge image into several small blocks, and count the connection of edge pixels in each small block. If the edge pixels in a small block form a continuous curve or line segment, the score of the small block is 1; otherwise, the score is 0. Calculate the average of all small block scores as the edge integrity score. High-frequency energy ratio: Perform Fourier transform on the image to convert the image from the spatial domain to the frequency domain. In the frequency domain, the high-frequency part corresponds to the details and edge information of the image. Calculate the ratio of high-frequency energy to total energy H. The specific process is as follows: First calculate the total energy of the high-frequency area in the frequency domain image , and then calculate the total energy of the entire frequency domain image ,but ; The higher the high-frequency energy ratio H, the richer the detail information contained in the image; Signal-to-Noise Ratio: Calculate the signal-to-noise ratio of the image; Assign corresponding weights to each image quality factor, calculate the comprehensive quality score of the image by weighted summation method, and compare the obtained comprehensive quality score with the preset quality score threshold. If the comprehensive quality score is greater than the quality score threshold, the image quality is judged to meet the requirements. Otherwise, the image is judged to have quality problems and further analyze the causes, including checking the transmission line, camera equipment or image acquisition environment; The early warning unit issues early warnings in different ways depending on the severity and type of the abnormality.

2. The shield machine operation and maintenance management system based on data analysis according to claim 1 is characterized in that: The execution process of the early warning unit is as follows: When the verification unit determines that the tool is abnormal, the early warning unit is activated immediately; different early warning methods are used according to the severity and type of the abnormality; For minor abnormalities, the operating room display screen will warn: a striking warning pop-up window will pop up on the display screen in the shield machine operating room. The pop-up window is displayed with a yellow background. The pop-up window content includes tool number, abnormal type, specific value of wear, current temperature rise value, and vibration amplitude deviation value detailed information, so that the operator can understand the abnormal situation of the tool; at the same time, the position of the tool where the abnormality occurs is marked with a dynamic flashing icon on the main interface of the display screen, and the abnormal prompt information is displayed next to the icon; Parameter adjustment suggestion prompt: Based on the abnormal situation analysis results, specific excavation parameter adjustment suggestions are provided on the display screen. The adjustment suggestions are presented in the form of a combination of text descriptions and dynamic charts; the tool status data is automatically updated at preset time intervals; For moderate abnormalities, the sound and light alarm in the operation room is activated: the sound and light alarm is triggered in the shield machine operation room, the alarm sound is an intermittent buzzer, and the light is a yellow flashing light; the sound and light alarm lasts for 10-15 seconds, and it is triggered again after an interval of 5 seconds; when the sound and light alarm is triggered, the display screen in the operation room switches to a full-screen display of the tool abnormality specific information page, which contains all the information for minor abnormalities, and also displays the vibration frequency deviation value, wear area parameter, and the duration of the current abnormal state; SMS notification sending: SMS notifications are automatically sent to on-duty technicians and on-site foremen on the ground. The SMS content includes the shield machine number, tool abnormality location, abnormality type, quantitative indicators of the abnormality degree, and currently recommended measures; SMS sending adopts priority setting, which is higher than work notifications but lower than emergency notifications; every 10 minutes, an updated tool abnormality information SMS is automatically sent until the abnormal situation is resolved or upgraded to a severe abnormality; For severe abnormalities, strong sound and light alarms: A strong sound and light alarm is activated in the shield machine operation room. The alarm sound is a continuous sharp high-pitched sound, and the light is a red flashing strong light. The sound and light alarm continues until the operator manually confirms that the alarm has been received or the abnormal situation has been handled; at the same time, all non-critical equipment in the operation room, including some auxiliary lighting and air conditioning, is automatically turned off to concentrate power supply to critical systems; a red warning sign and text prompts are displayed on the display screen, and the same content is repeatedly played in the form of voice broadcast; Multi-channel emergency notification: Immediately send emergency SMS notifications to ground engineering management personnel, equipment maintenance team leaders and relevant experts. The SMS content describes the severe abnormality of the tool, including all relevant quantitative data and the consequences, and requires relevant personnel to quickly organize forces to formulate maintenance plans and response measures; at the same time, send a detailed abnormality report to the above-mentioned personnel via email, and the report provides historical data trend charts and image analysis results of tool abnormalities in the form of attachments; automatically dial the preset emergency contact number and notify relevant personnel to handle the abnormality in the form of voice broadcast; Automatic shutdown preparation: Send instructions to the shield machine's control system to prepare for the automatic shutdown procedure; before shutdown, gradually reduce the shield machine's excavation speed to the lowest safe speed; at the same time, adjust the cutter head speed to the lowest stable speed; during the shutdown process, continuously monitor the tool status and shield machine parameters; once the shutdown is completed, lock some operating functions of the shield machine, including prohibiting the restart of the excavation program and limiting the rotation of the cutter head to prevent misoperation.

3. The shield machine operation and maintenance management system based on data analysis according to claim 2 is characterized in that: The severity assessment process of tool anomalies in the early warning unit is as follows: Obtain the parameters preprocessed by the data processing and analysis unit and perform preliminary analysis, including: Wear index: Tool residual thickness measurement: Regularly measure the tool's residual thickness. Suppose the tool's initial thickness is G0, and the currently measured residual thickness is G1, then the wear amount is G=G0-G1. Wear area calculation: For tools with irregular wear shapes, the area of ​​the wear area is calculated through image analysis technology. The tool surface is photographed to obtain a high-resolution image. The wear area boundary is identified using an image processing algorithm, and then the wear area YG is calculated. After normalizing the wear amount G and the wear area YG, the wear amount G and the wear area YG are used as the two right-angled sides of the right triangle, and the remaining side is connected to establish a right triangle. The area of ​​the right triangle is calculated and recorded as the wear evaluation value. Set different wear evaluation value intervals according to tool materials and design requirements; determine the severity of wear based on the wear evaluation value intervals; Vibration characteristic parameters: Vibration amplitude analysis: The amplitude of the vibration signal is calculated through the tool vibration data, and the average or peak value of the vibration amplitude in the preset time period is used to measure it. The deviation from the average or peak value of the normal vibration amplitude is used to calculate the vibration deviation value; Vibration frequency change: Perform spectrum analysis on the vibration signal to obtain the vibration frequency components, and calculate the vibration frequency deviation value through the vibration frequency components, which is recorded as the frequency deviation value; After the obtained vibration deviation value and frequency deviation value are normalized, the bottom circle is drawn with the frequency deviation value as the bottom circle radius, and a cone model is established with the vibration deviation value as the height, and the volume of the cone model is calculated and recorded as the vibration balance value; a number of vibration balance value intervals are preset according to the vibration balance value, and the preset vibration balance value intervals correspond to different vibration abnormality levels, and the vibration abnormality level is judged by the vibration balance value; Temperature change degree: Tool temperature rise value: The temperature of the tool during operation is measured by a temperature sensor installed near the tool, while the ambient temperature is recorded to calculate the temperature rise value; Temperature change rate: calculate the rate of change of tool temperature per unit time; The obtained temperature rise value and temperature change rate are calibrated as ws and wb respectively, and after normalization, they are entered into the following formula: To get the temperature estimate ZO, where is a correction coefficient; the obtained temperature value ZO is used as a measure of the degree of temperature change; the obtained temperature value ZO is compared with a plurality of preset temperature value intervals, and different temperature anomaly levels are set corresponding to the plurality of temperature value intervals, and the temperature anomaly level is determined by comparison; According to the importance of each indicator on the degree of tool abnormality, weights are assigned to the wear evaluation value, vibration balance value and temperature evaluation value. For each indicator, a score is given based on the comparison between its quantitative value and the set threshold. The comprehensive abnormality degree score is then calculated by the weighted summation method. The abnormality level is divided into slight abnormality, moderate abnormality and severe abnormality according to the comprehensive abnormality degree score.

4. The shield machine operation and maintenance management system based on data analysis according to claim 1 is characterized in that: The specific operation steps of the data acquisition unit are as follows: Install a variety of sensors on the cutter of the shield machine, including pressure sensors, vibration sensors, temperature sensors and cutter displacement sensors; The pressure sensor is used to detect the size and change of the cutting force that the tool bears during the process of cutting rock and soil; The vibration sensor can capture the vibration frequency and amplitude information of the tool caused by wear and collision; The temperature sensor monitors the temperature changes of the tool when it is working; The tool displacement sensor can measure the displacement of the tool in different directions to determine whether there is any abnormality; The sensor transmits the data collected in real time to the data transmission unit.

5. The shield machine operation and maintenance management system based on data analysis according to claim 1 is characterized in that: The specific operation steps of the data transmission unit are as follows: A hybrid wired and wireless transmission network is used to ensure that data can be transmitted from the data acquisition unit to the data processing and analysis unit. Inside the shield machine, a wired network is used to connect sensors and data transmission nodes to ensure the stability and anti-interference of data transmission. Wireless network is used for data transmission between the shield machine and the ground monitoring center to achieve remote and rapid data transmission.

6. The shield machine operation and maintenance management system based on data analysis according to claim 1 is characterized in that: The specific operation steps of the data processing and analysis unit are as follows: After receiving the raw data from the data transmission unit, data preprocessing is performed, including data cleaning, denoising, and data format conversion operations to remove outliers and interference factors in the data; Data analysis algorithms are used to conduct in-depth data mining and feature extraction. Specifically, wavelet analysis algorithms are used to perform time-frequency analysis on vibration data and extract vibration characteristics of different frequency bands. Principal component analysis algorithms are used to reduce the dimensionality of multi-dimensional sensor data and extract key feature vectors to reduce the complexity of data processing and highlight the characteristic information of the tool working status.

7. The shield machine operation and maintenance management system based on data analysis according to claim 1 is characterized in that: The specific operation steps of the abnormality detection unit are as follows: Based on the feature data extracted by the data processing and analysis unit, the anomaly detection unit builds an intelligent anomaly detection model, which specifically adopts a deep learning algorithm and uses a combination of a convolutional neural network and a long short-term memory network. The convolutional neural network is used to extract the spatial features of the data and identify the local feature patterns in the sensor data. The long short-term memory network can learn the change law of the tool working state over time in processing time series data, thereby capturing the development trend of the tool abnormal state. By inputting data on the normal and abnormal working conditions of known tools into the model for training, the model can automatically learn the characteristic patterns of the tools under different working conditions and establish corresponding discrimination criteria; when receiving new tool data, the model determines whether the tool is in an abnormal state and determines the type of abnormality, including wear, fracture, and eccentric wear.

Images