An infrared intelligent analysis system for ethylene cracking furnace tube

Abstract

The application discloses an ethylene cracking furnace tube infrared intelligent analysis system, which comprises, which are connected in sequence, an image acquisition module, a temperature monitoring module, a coking diagnosis module and a health management module; the image acquisition module acquires infrared images of cracking furnace tubes and burners by using an infrared thermal imaging online monitoring system, and establishes a cracking furnace tube infrared image database; the temperature monitoring module extracts and records the outer surface temperature of the ethylene cracking furnace tube from the processed infrared images of the ethylene cracking furnace tube by using an infrared image processing and recognition method; the coking diagnosis module combines the technologies of artificial intelligence and data mining, establishes a coking degree diagnosis and prediction model of the ethylene cracking furnace tube, and realizes early warning and targeted tracking of the coking abnormal conditions; the health management module carries out health evaluation of the cracking furnace tube according to the results obtained by the coking degree diagnosis and prediction model of the ethylene cracking furnace tube, and provides maintenance decision support. The application can realize integrated management of cracking furnace tube detection, diagnosis and maintenance.

Description

TECHNICAL FIELD

[0001] The present application relates to a kind of temperature detection of ethylene cracking furnace tube.The particularly relates to a kind of ethylene cracking furnace tube infrared intelligent analysis system. BACKGROUND

[0002] Ethylene is the most basic raw material of petrochemical industry, and the world has made ethylene production as one of the important indicators to measure the development level of a country's petrochemical industry. However, in the process of producing ethylene by thermal cracking method, the cracking furnace tube inevitably produces carburizing and coking, which will greatly affect the production efficiency of ethylene enterprises. Among the many causes of cracking furnace tube coking, the metal temperature of the outer surface of the cracking furnace tube is one of the most important factors. Therefore, how to accurately measure the temperature of the outer surface of the cracking furnace tube, and then accurately diagnose and predict the coking degree of the cracking furnace tube, has become one of the key problems to ensure the safe operation of the cracking furnace, improve the production efficiency of the cracking furnace and make the ethylene industry intelligent.

[0003] At present, the temperature measurement technology of the outer surface of the cracking furnace tube mainly includes contact type and non-contact type. The contact type temperature measurement method is to put the temperature measurement element directly into the measured object or its temperature field during temperature measurement. The most common contact type temperature measurement element is thermocouple temperature sensor. Although this method has high measurement accuracy, the thermocouple is in close contact with the measured object, which will affect the temperature field distribution of the measured object. Moreover, the thermocouple is easily damaged or has temperature drift in high temperature environment, and once damaged, it can only be replaced by stopping the furnace for repair, which often causes blind burning due to the inability to replace it in time. In addition, since the thermocouple is located at the bottom of the furnace tube, it cannot effectively monitor the temperature and the highest temperature of the radiant section of the cracking furnace. In terms of non-contact temperature measurement method, common methods include infrared temperature measurement gun positioning temperature measurement method, infrared thermal imaging method, multi-spectral temperature measurement method, etc. Among them, the most widely used method is manual holding of non-contact infrared temperature measurement gun, aiming and measuring through the observation window of the cracking furnace. However, the temperature around the observation window of the cracking furnace is very high, and it is difficult to measure by manual aiming. Moreover, this method has the disadvantages of low measurement accuracy, strong randomness, limited number of furnace tubes measured at one time, etc.

[0004] In recent years, some cracking furnaces have begun to use infrared thermal imaging for online monitoring of the furnace tube, but the functions are mainly visual display of the surface temperature of the furnace tube and over-temperature alarm, etc., and cannot effectively predict the coking trend and evaluate the service life of the furnace tube. During the coking process, there is a lack of effective process tracking means, which causes incomplete coking and aggravates the coking of the furnace tube after starting, and in severe cases, it can cause the furnace tube to be blocked. The burner failure cannot be found in time, the COT of a single furnace tube is too high, which causes the metal of the inner surface of the furnace tube to be over-oxidized, and the furnace tube is seriously coked.

[0005] The prior art detects the surface temperature of the ethylene cracking furnace mainly by measuring temperature data and distance data through an intelligent temperature measuring device, and can only detect the average temperature of the whole furnace tube, and cannot obtain the temperature value of the specified point of the furnace tube. SUMMARY

[0006] The technical problem to be solved by the present application is to overcome the shortcomings of the prior art, and to provide an ethylene cracking furnace tube infrared intelligent analysis system capable of ensuring the safe operation of the cracking furnace and improving the production efficiency of the cracking furnace.

[0007] The technical solution adopted by the present application is: an ethylene cracking furnace tube infrared intelligent analysis system, comprising, in sequence, an image acquisition module, a temperature monitoring module, a coking diagnosis module and a health management module; wherein:

[0008] The image acquisition module acquires the infrared images of the cracking furnace tube and the burner by using an infrared thermal imaging online monitoring system, and establishes a cracking furnace tube infrared image database;

[0009] The temperature monitoring module adopts an infrared image processing and recognition method to extract and record the outer surface temperature of the ethylene cracking furnace tube from the processed infrared images of the ethylene cracking furnace tube;

[0010] The coking diagnosis module combines artificial intelligence and data mining technology to establish a coking degree diagnosis and prediction model of the ethylene cracking furnace tube, and realizes early warning and targeted tracking of coking abnormal conditions;

[0011] The health management module performs health assessment of the cracking furnace tube based on the results obtained by the coking degree diagnosis and prediction model of the ethylene cracking furnace tube, and provides maintenance decision support.

[0012] The ethylene cracking furnace tube infrared intelligent analysis system of the present application can realize integrated management of cracking furnace tube detection, diagnosis and maintenance. The present application acquires the infrared images of the cracking furnace by using an infrared thermal imaging camera, can obtain the temperature distribution map of the furnace tube under any specified viewing angle, and can query the surface temperature of the furnace tube at any point. At the same time, the present application can monitor and record the temperature data of the furnace tube in real time, set up a high temperature alarm system, and guide relevant technical personnel to operate the cracking furnace in a refined manner. The present application uses AI technology to analyze the infrared video images of the cracking furnace in depth, accurately diagnoses and predicts the coking degree of the cracking furnace tube, and then establishes an AI infrared analysis-based coking period automatic optimization system and a furnace tube life assessment system, so as to ensure the safe operation of the cracking furnace, improve the production efficiency of the cracking furnace, and make the ethylene industry intelligent. The present application combines artificial intelligence and data mining technology to construct an intelligent health management system for the cracking furnace tube, and uses fault prediction and health management technology to predict the faults of the furnace tube, automatically optimize the coking, and assess the life of the furnace tube, so as to realize the detection, positioning, health assessment, fault prediction and maintenance decision of the furnace tube. Attached Figure Description

[0013] Figure 1 This is a block diagram of the infrared intelligent analysis system for ethylene cracking furnace tubes according to the present invention;

[0014] Figure 2 This is a schematic diagram of the LSTM neuron structure. Detailed Implementation

[0015] The following describes in detail an infrared intelligent analysis system for ethylene cracking furnace tubes according to the present invention, with reference to embodiments and accompanying drawings.

[0016] like Figure 1 As shown, an infrared intelligent analysis system for ethylene cracking furnace tubes according to the present invention includes, in sequence: an image acquisition module 1, a temperature monitoring module 2, a coking diagnosis module 3, and a health management module 4; wherein:

[0017] The image acquisition module 1 uses an infrared thermal imaging online monitoring system to acquire infrared images of the pyrolysis furnace tubes and burners, establishing an infrared image database of the pyrolysis furnace tubes. The temperature monitoring module 2 uses infrared image processing and recognition methods to extract and record the outer surface temperature of the ethylene pyrolysis furnace tubes from the processed infrared images. The coking diagnosis module 3 combines artificial intelligence and data mining techniques to establish a diagnostic and predictive model for the degree of coking in the ethylene pyrolysis furnace tubes, enabling early warning and targeted tracking of abnormal coking conditions. The health management module 4 performs a health assessment of the furnace tubes based on the results obtained from the diagnostic and predictive model for the degree of coking in the ethylene pyrolysis furnace tubes, providing maintenance decision support. Wherein:

[0018] 1. The image acquisition module 1 uses an infrared thermal imaging online monitoring system to acquire infrared images of the pyrolysis furnace tubes and burners, and establishes an infrared image database of the pyrolysis furnace tubes, specifically including:

[0019] (1.1) Infrared images of multiple ethylene cracking furnace tubes from different angles are acquired in real time using infrared thermal imaging equipment to continuously monitor the changes and distribution of surface temperature of the furnace tubes;

[0020] (1.2) Preprocess the infrared image of the ethylene cracking furnace tube, including using image enhancement algorithms to improve image contrast, remove noise, extract detail information, and improve image clarity;

[0021] (1.3) Based on the infrared images of the ethylene cracking furnace tubes from different perspectives obtained by processing, a fault database of the cracking furnace tube infrared images is established for subsequent temperature monitoring, coking diagnosis and health management of the furnace tubes.

[0022] 2. The temperature monitoring module 2 uses infrared image processing and recognition method to extract and record the ethylene cracking furnace tube outer surface temperature from the processed infrared image of ethylene cracking furnace tube, specifically including:

[0023] (2.1) The algorithm of image segmentation is used to identify the shape of the cracking furnace tube, and the spatial distribution of the camera is combined to locate and identify multiple furnace tubes under different viewing angles;

[0024] (2.2) The infrared images of each furnace tube under different viewing angles are respectively processed by weighted average to reduce the influence of flame or heat flow on the infrared image of the furnace tube, and the temperature of the furnace tube is more accurately represented;

[0025] (2.2) Map the gray value in the infrared image to the actual surface temperature, get the furnace tube temperature distribution map in the spatial dimension, and facilitate real-time query of the furnace tube surface temperature at any point; Get the surface temperature time series of any area of the furnace tube in the time dimension, which is convenient for viewing temperature changes and intuitively mastering the actual running state of the furnace tube at the concerned position.

[0026] 3. The coking diagnosis module 3 combines artificial intelligence and data mining technology to establish an ethylene cracking furnace tube coking degree diagnosis and prediction model, realizes early warning and targeted tracking of coking abnormal conditions, and is firstly the induction analysis of the degradation rule of the detected furnace tube outer surface temperature time series. Since the degradation data has strong dynamics and randomness, the degradation model is divided into trend item and random item two parts, and then the failure threshold and failure mode of the furnace tube coking are matched with the degradation model. Then, based on this, the online data is analyzed, the fault prediction is carried out combined with artificial intelligence algorithm, the furnace tube coking degree diagnosis model is established, and the occurrence time and probability of the furnace tube fault are predicted. Specifically including:

[0027] (3.1) Time series decomposition of furnace tube surface temperature data,

[0028] The STL (Seasonal and Trend decomposition using Loess) decomposition method is used to process time series decomposition. Specifically, a robust local weighted regression method is used to decompose the time series into a trend item, a periodic item and a residual item, which are used to measure the strength of the trend item and the periodic item in the time series. The trend item reflects a trend or state of continuous development and change over a long period of time. The periodic item reflects the regular variation of the development level of the phenomenon. The residual item reflects the influence of numerous accidental factors on the time series, such as the Wiener process. The local weighted regression process and the robust process of the robust local weighted regression method are implemented in the inner loop and the outer loop of the STL decomposition, respectively. In the inner loop, the periodic item is obtained by removing the trend item, and the trend item is obtained by removing the periodic item. After convergence, the trend item and the periodic item are smoothed by local weighted regression. In the outer loop, the robust weight value of each sample point in the sequence is calculated, and the neighborhood weight value is multiplied by the robust weight value when local weighted regression is performed in the inner loop.

[0029] (3.2) Diagnose and predict the degree of coking of the furnace tube, including:

[0030] (3.2.1) Divide the coking degree of the furnace tube into four levels: normal, mild coking, moderate coking and severe coking;

[0031] The four levels are divided according to the surface temperature of the furnace tube, and the surface temperature of the furnace tube is T. The specific division is as follows:

[0032] Normal: 800℃≤T≤860℃;

[0033] Mild coking: 860℃<T≤920℃;

[0034] Moderate coking: 920℃<T<1100℃;

[0035] Severe coking: 1100℃≤T.

[0036] (3.2.2) Randomly divide 70% of the time series data of the surface temperature of the furnace tube into a training set and 30% into a test set; use the Adam optimization method to train the long short-term memory neural network (LSTM) through the training set, and test the performance of the trained long short-term memory neural network on the test set;

[0037] (3.2.3) Use the trained long short-term memory neural network to diagnose the coking degree of the furnace tube during the operation cycle of the furnace tube, and draw a curve of the coking degree with respect to time to predict the development trend of the coking degree of the furnace tube.

[0038] LSTM is a time series convolutional neural network derived from recurrent neural networks (RNN), which has certain long-distance time series data mining capabilities. In the standard RNN, there is a chain form of repeated neural network modules, but this module only has a very simple structure, such as a tanh layer. LSTM is also such a structure, but the repeated module has a different structure, unlike a single neural network layer, and there are four special ways to interact in LSTM, as shown in Figure 2 .

[0039] The LSTM network introduces a gating mechanism to control the path of information transmission: input gate i t , forget gate f t and output gate o t .

[0040] ①The forget gate f t controls how much information needs to be forgotten from the internal state c t-1 of the last moment;

[0041] ②The input gate i t controls how much information needs to be saved in the candidate state of the current moment;

[0042] ③The output gate o t controls how much information needs to be output to the external state h t from the internal state c t of the current moment.

[0043] When f t = 0, i t = 1, the memory cell clears the historical information and writes the candidate state vector . But at this time, the memory cell c t is still related to the historical information of the last moment; when f t = 1, i t = 0, the memory cell will copy the content of the last moment, without writing new information.

[0044] The "gate" in the LSTM network is a "soft gate" with a value between (0, 1), indicating that information is allowed to pass at a certain rate. The calculation method of the three gates is:

[0045] i t = σ(W i x t + U i h t-1 + b i )

[0046] f t =σ(W f x t +U f h t-1 +b f )

[0047] o t =σ(W0x t +U0h t-1 +b o )

[0048] Where σ(·) is the Logistic function, and its output interval is (0,1), x t h is the input at the current moment. t-1 This represents the external state at the previous moment.

[0049] The calculation process for the cyclic cell structure of an LSTM network is as follows:

[0050] ① First, use the external state h from the previous time step. t-1 and the input x at the current time t The three gates and candidate states were calculated.

[0051] ② Combining the forgetting gate f t and input gate i t To update memory unit c t ;

[0052] ③ Combine with output gate o t The information of the internal state is passed to the external state h. t .

[0053] Through LSTM recurrent units, the entire network can establish long-distance temporal dependencies. The calculation formula for LSTM can be expressed as:

[0054]

[0055]

[0056] h t =o t ⊙tanh(c t )

[0057] Where, x t The input at the current moment is W, and the network parameters are b.

[0058] The training steps for an LSTM network are as follows:

[0059] ① Input the data features at time t into the input layer, and output the results after passing through the activation function.

[0060] ②The input layer output result, the hidden layer output at t-1 moment and the information stored by the memory unit at t-1 moment are input into the nodes of the LSTM structure, and the data is output to the next hidden layer or the output layer through the processing of the input gate, the output gate, the forgetting gate and the memory unit.

[0061] ③The output of the LSTM structure node is output to the output layer neuron, and the output result is output.

[0062] ④Error back propagation, update each weight.

[0063] 4. The health management module 4 performs health assessment on the results obtained by the ethylene cracking furnace tube coking degree diagnosis and prediction model, and provides maintenance decision support, which is a health state assessment model established for the furnace tube in combination with the cracking furnace tube coking degree diagnosis and prediction model. The historical coking times and maintenance conditions and other test information are fused to study the preventive maintenance guarantee strategy, establish a decision model for maintenance cycle and maintenance opportunity, and provide intelligent push for maintenance detection. Specifically, it includes:

[0064] (4.1) Construct a furnace tube health state assessment system, including:

[0065] (4.1.1) Determine the comment set of the furnace tube health state assessment system, specifically divide the health state of the furnace tube into four levels: healthy, attention, deterioration and disease; wherein:

[0066] The healthy level indicates that the overall health condition of the furnace tube is good, the furnace tube has no coking or only one area of mild coking, and the ethylene cracking function is normal;

[0067] The attention level indicates that the overall health condition of the furnace tube is general, the furnace tube has a mild coking area or one area of moderate coking, the ethylene cracking function is normal, and regular maintenance should be performed;

[0068] The deterioration level indicates that the overall health condition of the furnace tube is poor, the furnace tube has ≥2 moderate coking areas, and one area of severe coking, the ethylene cracking function is damaged, and timely maintenance is needed;

[0069] The disease level indicates that the overall health condition of the furnace tube is poor, there are ≥2 severe coking areas, the ethylene cracking function is lost, and emergency maintenance or replacement is needed;

[0070] (4.1.2) Determine the element set of the furnace tube health state assessment system, specifically the evaluation elements that can reflect the health condition of the furnace tube, including the coking degree of the furnace tube, the pressure of the furnace tube, the inlet temperature of the furnace tube and the outlet temperature of the furnace tube;

[0071] (4.1.3) Determine the weight set of the furnace tube health state evaluation system, specifically determine the weights of the four evaluation elements of the furnace tube coking degree, the furnace tube pressure, the furnace tube inlet temperature and the furnace tube outlet temperature by the analytic hierarchy process, and constitute the weight set A={a1, a2, a3, a4}, wherein a1, a2, a3, a4 represent the weights corresponding to the furnace tube coking degree evaluation element, the furnace tube pressure evaluation element, the furnace tube inlet temperature evaluation element and the furnace tube outlet temperature evaluation element respectively;

[0072] (4.1.4) The fuzzy comprehensive evaluation method with semi-trapezoidal distributed membership function is used for furnace tube health evaluation, specifically including:

[0073] Firstly, the semi-trapezoidal distributed membership functions of the four comment sets of health, attention, deterioration and disease are constructed, respectively represented as and

[0074] Then, the four evaluation elements of the furnace tube coking degree, the furnace tube pressure, the furnace tube inlet temperature and the furnace tube outlet temperature are substituted into the semi-trapezoidal distributed membership functions of the four comment sets to determine the degree of the elements belonging to each evaluation grade, i.e. the membership degree, denoted as r ij , r ij represents the membership degree of the i-th evaluation element corresponding to the j-th evaluation grade, i=1, 2, 3, 4; j=1, 2, 3, 4;

[0075] Finally, the weighted average type fuzzy operator is used for calculation, and the membership degree b j of the furnace tube corresponding to the j-th evaluation grade is obtained comprehensively, and the formula is The health comprehensive evaluation result of the furnace tube is finally determined according to the maximum membership degree principle; a i is the weight of the i-th evaluation element.

[0076] (4.2) Furnace tube health management based on industrial internet, including:

[0077] (4.2.1) Whole life cycle visualization: Construct a digital file of the furnace tube, including static attribute data of the furnace tube model, size and measuring point position, and dynamic attribute data of the history repair, maintenance record and state trend of the furnace tube, collect the whole process management data records of the furnace tube from installation, operation, change, repair, maintenance and scrap, and form a complete equipment history file of static data and dynamic data;

[0078] (4.2.2) State monitoring: Use the maximum threshold index of the furnace tube surface temperature and the temperature difference information index to find and alarm the abnormal position in time, record the abnormal position information and abnormal state information; and record the temperature in real time for the furnace tube part area which has appeared abnormal or needs to be focused on, and use the historical temperature information to perform data statistics, analysis and drawing, for example, draw the temperature-time sequence change graph of the key position, and track the temperature state change of the key position in a targeted manner;

[0079] (4.2.3) Maintenance management: According to the health state of the furnace tube, combined with the comprehensive performance, average failure-free operation time, average maintenance time, preventive maintenance work completion rate, furnace tube maintenance cost, furnace tube maintenance man-hours, fault cause of the furnace tube, realize comprehensive comparison and analysis in multiple dimensions of time, equipment and personnel, support periodic repair, condition-based repair, spot repair, after-repair, outsourcing repair, provide intelligent push for maintenance detection personnel, and assist in the diagnosis and maintenance of the furnace tube.

Claims

1. An infrared intelligent analysis system for ethylene cracking furnace tubes, characterized in that, It includes, in sequence: an image acquisition module (1), a temperature monitoring module (2), a coking diagnosis module (3), and a health management module (4); wherein: The image acquisition module (1) uses an infrared thermal imaging online monitoring system to acquire infrared images of the pyrolysis furnace tubes and burners, and establishes an infrared image database of the pyrolysis furnace tubes; The temperature monitoring module (2) uses an infrared image processing and recognition method to extract and record the outer surface temperature of the ethylene cracking furnace tube from the processed infrared image of the ethylene cracking furnace tube. The coking diagnosis module (3) combines artificial intelligence and data mining technologies to establish a diagnostic and prediction model for the degree of coking in the furnace tubes of the ethylene cracking furnace, so as to realize early warning and targeted tracking of abnormal coking conditions. The health management module (4) performs a health assessment of the furnace tubes based on the results obtained from the ethylene cracking furnace tube coking degree diagnosis and prediction model, and provides maintenance decision support; including: (4.1) Construct a furnace tube health status assessment system, including: (4.1.1) Determine the evaluation criteria for the furnace tube health status assessment system, specifically classifying the furnace tube health status into four levels: healthy, attentive, deteriorating, and diseased; the four levels of health, attentive, deteriorating, and diseased are as follows: The health rating indicates that the overall health of the furnace tubes is good, there is no coking in the furnace tubes or only one area has slight coking, and the ethylene cracking function is normal. The "Note" rating indicates that the overall health of the furnace tubes is generally good, with mild coking areas or moderate coking in one area. The ethylene cracking function is normal, and regular maintenance is required. The deterioration level indicates that the overall health of the furnace tubes is poor, with ≥2 areas of moderate coking and 1 area of ​​severe coking, impairing the ethylene cracking function and requiring timely maintenance. The disease level indicates that the overall health of the furnace tubes is poor, with ≥2 severely coking areas, resulting in loss of ethylene cracking function, requiring urgent maintenance or replacement; (4.1.2) Determine the element set of the furnace tube health status assessment system, specifically the evaluation elements that can reflect the health status of the furnace tube, including the degree of coking of the furnace tube, furnace tube pressure, furnace tube inlet temperature and furnace tube outlet temperature; (4.1.3) Determine the weight set of the furnace tube health status assessment system. Specifically, use the analytic hierarchy process (AHP) to determine the weights corresponding to the four evaluation elements: furnace tube coking degree, furnace tube pressure, furnace tube inlet temperature, and furnace tube outlet temperature. This weight set is denoted as […]. Where a1, a2, a3, and a4 represent the weights of the furnace tube coking degree evaluation element, furnace tube pressure evaluation element, furnace tube inlet temperature evaluation element, and furnace tube outlet temperature evaluation element, respectively. (4.1.4) The fuzzy comprehensive evaluation method using semi-trapezoidal distributed membership functions is adopted for furnace tube health assessment; specifically including: First, construct semi-trapezoidal distributed membership functions for four comment sets: health, attention, deterioration, and disease, denoted as follows: , , and ; Then, the four evaluation elements—furnace tube coking degree, furnace tube pressure, furnace tube inlet temperature, and furnace tube outlet temperature—are substituted into the semi-trapezoidal distributed membership functions of the four evaluation sets to determine the degree to which an element belongs to each evaluation level, i.e., the membership degree, denoted as . , This represents the membership degree of the i-th evaluation element to the j-th evaluation level, where i=1,2,3,4; j=1,2,3,4; Finally, a weighted average fuzzy operator is used to calculate and synthesize the membership degree of the furnace tube corresponding to the j-th evaluation level. The formula is The final comprehensive health assessment result of the furnace tubes is determined according to the principle of maximum membership; a i Let be the weight of the i-th evaluation element; (4.2) Furnace tube health management based on the Industrial Internet, including: (4.2.1) Full life cycle visualization: Construct a digital archive of furnace tubes, including: static attribute data of furnace tube model, size and measuring point location, as well as dynamic attribute data of furnace tube historical repair, maintenance records and status trends, and collect the full management data records of furnace tubes from installation, operation, changes, repair, maintenance and scrapping, forming a complete equipment history archive with static and dynamic data; (4.2.2) Status monitoring: The maximum threshold index of furnace tube surface temperature and temperature difference information index are used to detect and alarm the location of abnormality in a timely manner, and the location information and status information of abnormality are recorded; the temperature of furnace tube areas that have experienced abnormalities or require special attention is recorded in real time, and historical temperature information is used to perform data statistics, analysis and plotting, and to track the temperature status changes of key locations in a targeted manner. (4.2.3) Maintenance and repair management: Based on the health status of the furnace tubes, combined with the furnace tubes' comprehensive performance, mean time between failures, mean time to repair, completion rate of preventive maintenance work, furnace tube maintenance costs, furnace tube maintenance hours, and causes of failure, comprehensive comparison and analysis are achieved in multiple dimensions such as time, equipment, and personnel. It supports periodic maintenance, condition maintenance, spot inspection maintenance, post-event maintenance, and outsourced maintenance, and provides intelligent push for maintenance and testing personnel to assist in the diagnosis and repair of furnace tubes.

2. The infrared intelligent analysis system for ethylene cracking furnace tubes according to claim 1, characterized in that, The image acquisition module (1) uses an infrared thermal imaging online monitoring system to acquire infrared images of the pyrolysis furnace tubes and burners, and establishes an infrared image database of the pyrolysis furnace tubes, including: (1.1) Infrared images of multiple ethylene cracking furnace tubes from different angles are acquired in real time using infrared thermal imaging equipment to continuously monitor the changes and distribution of surface temperature of the furnace tubes; (1.2) Preprocess the infrared image of the ethylene cracking furnace tube, including using image enhancement algorithms to improve image contrast, remove noise, extract detail information, and improve image clarity; (1.3) Based on the infrared images of the ethylene cracking furnace tubes from different perspectives obtained by processing, a fault database of infrared images of cracking furnace tubes is established for subsequent temperature monitoring, coking diagnosis and health management of the furnace tubes.

3. The infrared intelligent analysis system for ethylene cracking furnace tubes according to claim 1, characterized in that, The temperature monitoring module (2) uses infrared image processing and recognition methods to extract and record the outer surface temperature of the ethylene cracking furnace tube from the processed infrared image of the tube, including: (2.1) The shape of the pyrolysis furnace tubes is identified by using image segmentation algorithms, and multiple furnace tubes are located and identified from different perspectives by combining the spatial distribution of the cameras. (2.2) The infrared images of each furnace tube under different viewing angles are processed by weighted average to reduce the influence of flame or heat flow on the infrared images of the furnace tube and to more accurately represent the temperature of the furnace tube. (2.2) The gray values ​​in the infrared image are mapped to the actual surface temperature. The furnace tube temperature distribution map is obtained in the spatial dimension, which makes it easy to query the surface temperature of the furnace tube at any point in real time. The surface temperature time series of any area of ​​the furnace tube is obtained in the time dimension, which makes it easy to view the temperature change and intuitively grasp the actual operating status of the furnace tube at the location of interest.

4. The infrared intelligent analysis system for ethylene cracking furnace tubes according to claim 1, characterized in that, The coking diagnosis module (3) combines artificial intelligence and data mining technologies to establish a diagnostic and predictive model for the degree of coking in ethylene cracking furnace tubes, enabling early warning and targeted tracking of abnormal coking conditions, including: (3.1) Perform time series decomposition on the furnace tube surface temperature data. The STL decomposition method is used to decompose time series data. Specifically, a robust locally weighted regression method is used to decompose the time series into a trend term, a periodic term, and a residual term to measure the strength of the trend and periodic terms in the time series. The locally weighted regression process and the robustness process of the robust locally weighted regression method are nested within the inner and outer loops of the STL decomposition, respectively. The inner loop iterates through detrending the periodic term to obtain the trend term, and deperiodic regression to obtain the trend term. After convergence, the trend and periodic terms are obtained after smoothing by locally weighted regression. The outer loop calculates and updates the robust weight values ​​of each sample point in the sequence. Correspondingly, when performing locally weighted regression in the inner loop, the neighborhood weights need to be multiplied by the robust weight values. (3.2) Diagnose and predict the degree of coking in furnace tubes, including: (3.2.1) The degree of coking of furnace tubes is divided into four levels: normal, light coking, moderate coking and severe coking; (3.2.2) The time series data of furnace tube surface temperature is randomly divided into 70% as the training set and 30% as the test set. The Adam optimization method is used to train the long short-term memory neural network through the training set, and the performance of the trained long short-term memory neural network is tested on the test set. (3.2.3) The trained long short-term memory neural network is used to diagnose the degree of coking in the furnace tubes during the operation cycle of the pyrolysis furnace, and a curve of the degree of coking as a function of time is generated to predict the development trend of the degree of coking in the furnace tubes.

5. The infrared intelligent analysis system for ethylene cracking furnace tubes according to claim 4, characterized in that, The four levels described in section (3.2.1) are classified according to the surface temperature of the furnace tubes. Let the surface temperature of the furnace tubes be T. The specific classification is as follows: Normal: 800 0 C≤T≤860 0 C; Mild coking: 860 0 C <T≤920 0 C; Moderate coking: 920 0 C <T<1100 0 C; Severe coking: 1100 0 C≤T.

Images