A method and system for fault diagnosis and prediction of heating surface based on boiler flow field analysis

Abstract

The present application relates to the technical field of boiler heating surface fault diagnosis, and particularly relates to a heating surface fault diagnosis and prediction method and system based on boiler flow field analysis. The method comprises the following steps: collecting multi-dimensional data of the boiler; performing internal flow field analysis of the boiler; constructing a heating surface fault diagnosis model; coupling the flow field simulation results and the fault diagnosis model to construct a fault prediction model; extracting features and reducing dimensions; using a Gaussian mixture model to perform fault mode recognition; calculating a comprehensive fault score and performing risk classification. The present application significantly improves the accuracy of fault diagnosis and the early warning time by fusing flow field analysis, deep learning and multi-source data. Experimental results show that, compared with traditional methods, the fault detection rate of the present application is increased to 96.7%, the false positive rate is reduced to 3.2%, the early warning time is extended to 12.3 hours, and the diagnosis accuracy is 94.1%. The present application provides strong support for safe and efficient operation of the boiler, and has important engineering application value.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of boiler heating surface fault diagnosis, and in particular to a heating surface fault diagnosis and prediction method and system based on boiler flow field analysis. BACKGROUND

[0002] With the rapid development of industrial automation and intelligent manufacturing, the safe and stable operation of boilers as the core equipment of thermal systems is crucial to energy production and industrial processes. In recent years, data-driven fault diagnosis and prediction technologies have been widely applied in the field of boiler operation and maintenance. Traditional fault diagnosis methods mainly rely on expert experience and simple threshold judgment, which are difficult to cope with complex working condition changes and multi-source heterogeneous data. With the advancement of computational fluid dynamics (CFD) and machine learning technologies, fault diagnosis methods based on flow field analysis have gradually attracted attention. These methods can more accurately describe the working state and potential fault risk of the heating surface by constructing an internal flow field model of the boiler and combining real-time monitoring data. However, current flow field analysis methods are mostly limited to static simulation, making it difficult to capture dynamic characteristics during boiler operation, and lack comprehensive consideration of multi-scale, multi-physical field coupling effects.

[0003] There are still some deficiencies in existing boiler heating surface fault diagnosis and prediction technologies. First, most methods only focus on a single physical quantity (such as temperature, pressure), ignoring the comprehensive influence of flow field distribution, chemical reactions, and other factors on the state of the heating surface, resulting in biased diagnosis results. Second, traditional fault prediction models often use linear or simple nonlinear methods, which are difficult to effectively capture the complex dynamic characteristics and potential fault patterns of the boiler system. Third, existing systems generally lack explainability of the diagnosis results, making it difficult to provide intuitive and clear decision support for operation and maintenance personnel. In addition, most methods fail to effectively integrate historical fault data, expert knowledge, and physical models, limiting the accuracy and reliability of diagnosis and prediction. Finally, existing systems often show poor generalization ability and adaptability when facing new fault patterns or working condition changes. Therefore, it is urgent to develop an intelligent diagnosis and prediction method that can comprehensively consider flow field characteristics, multi-source data fusion, and nonlinear dynamic modeling to improve the accuracy, reliability, and adaptability of boiler heating surface fault diagnosis, and provide strong support for the safe and efficient operation of boilers. SUMMARY

[0004] In view of the problems existing in the prior art, the present application is proposed.

[0005] Therefore, the problem to be solved by the present application is that firstly, most methods only focus on a single physical quantity (such as temperature, pressure), ignoring the comprehensive influence of factors such as flow field distribution and chemical reaction on the state of the heating surface, resulting in deviation of the diagnosis result. Secondly, the traditional fault prediction model often uses linear or simple nonlinear methods, which is difficult to effectively capture the complex dynamic characteristics and potential fault modes of the boiler system. Thirdly, the existing system generally lacks explainability of the diagnosis result, which is difficult to provide intuitive and clear decision support for the operation and maintenance personnel. In addition, most methods fail to effectively integrate historical fault data, expert knowledge and physical models, limiting the accuracy and reliability of diagnosis and prediction. Finally, the existing system often shows poor generalization ability and adaptability when facing new fault modes or changes in working conditions.

[0006] To solve the above technical problems, the present application provides the following technical solutions:

[0007] In a first aspect, the embodiments of the present application provide a heating surface fault diagnosis and prediction method based on boiler flow field analysis, which comprises collecting boiler data, analyzing the internal flow field of the boiler, constructing a heating surface fault diagnosis model by capturing flow characteristics and heat transfer characteristics, and outputting historical data and environmental factors of fault occurrence.

[0008] A heating surface fault prediction model is constructed by coupling the flow field simulation results output by the boiler flow field analysis and the heating surface fault diagnosis model; and real-time monitoring data of fault prediction of the heating surface of the boiler are output.

[0009] The feature information obtained by flow field analysis is used to construct an intelligent diagnosis system in combination with real-time monitoring data, potential faults are identified through the intelligent diagnosis system, and preventive maintenance suggestions are provided.

[0010] As a preferred scheme of the heating surface fault diagnosis and prediction method based on boiler flow field analysis, the collecting of boiler data comprises flow field parameter data, heating surface state data, operating parameter data, environmental parameters, historical fault data, equipment state data, and maintenance records.

[0011] Flow field parameter data: flow velocity distribution, temperature distribution, pressure distribution, and flue gas composition, wherein the flue gas composition comprises O2, CO2, NO x ;

[0012] Heating surface state data: pipe wall temperature, pipe wall thickness, surface corrosion degree, and fouling condition.

[0013] Operating parameter data: boiler load, feedwater flow, steam parameter, fuel input quantity and property, wherein the steam parameter comprises temperature, pressure, and flow.

[0014] Environmental parameters: environmental temperature, humidity, and atmospheric pressure.

[0015] Historical failure data: past failure types, failure occurrence time, failure duration, failure impact range;

[0016] Equipment status data: vibration data, noise level, equipment running time;

[0017] Maintenance records: regular maintenance data, replacement parts information, cleaning records.

[0018] As a preferred scheme of the boiler flow field analysis-based heating surface fault diagnosis and prediction method of the present application, wherein: the collected boiler data is analyzed to analyze the internal flow field of the boiler, and the velocity distribution and temperature distribution of the collected fluid in the boiler are provided as basic data for the analysis of the heating surface of the boiler, and the internal flow field analysis of the boiler is constructed, which adopts the following expression:

[0019]

[0020] Where, ρ is the density of flue gas, calculated from flue gas composition data; v is the flow velocity distribution, a flow field parameter; P is the pressure distribution, a flow field parameter; μ is the viscosity of flue gas, calculated from fuel input and properties; c p is the specific heat capacity of flue gas, calculated from flue gas composition data; T is the temperature distribution, a flow field parameter; k is the thermal conductivity of flue gas, calculated from fuel input and properties; S is the heat source term, calculated from fuel input and boiler load data;

[0021] The flow field and temperature distribution analysis outputs detailed flow velocity and temperature distribution flow characteristics, and serves as an input for the state change of the heating surface, and outputs the heat transfer characteristics of the boiler heating surface state and corrosion prediction and the boiler heating surface state and fouling prediction through the change of the heating surface tube wall thickness with time, and provides key parameters for constructing the heating surface fault diagnosis model, and constructs the closing thickness change equation combining the flow field temperature and chemical reaction, which adopts the following expression:

[0022]

[0023] Where, W is the tube wall thickness, a heating surface state data; k c is the corrosion constant, calculated from past failure type data; E c is the corrosion activation energy, calculated from past failure type data; R is the gas constant; T is the temperature, the temperature distribution output in the last step; C corrosive is the corrosion substance concentration, a flue gas composition in the flow field parameter, which includes O2, CO2, NO x ; k s is the fouling constant, calculated from past failure type data; C scaling is the fouling substance concentration, a flue gas composition in the flow field parameter;

[0024] The flow characteristics and heat transfer characteristics are captured to construct a heating surface fault diagnosis model, and historical data and environmental factors of fault occurrence are output; the coupling results of the flow field, temperature distribution and pipe wall thickness change are taken as input variables, the flow field, heating surface state, operating parameters, environmental parameters, historical fault data and maintenance records are used to construct a heating surface fault diagnosis model, and real-time fault prediction is carried out, which adopts the following expression:

[0025]

[0026] Wherein, is the predicted fault state; X flow (t) is the flow field parameter, flow velocity distribution, temperature distribution, pressure distribution, flue gas composition; X heat (t) is the heating surface state parameter, pipe wall temperature, pipe wall thickness, surface corrosion degree, fouling condition; X run (t) is the operating parameter, boiler load, feed water flow, steam parameter, fuel input and nature; X env (t) is the environmental parameter, environmental temperature, humidity, atmospheric pressure; X hist (t) is the historical fault data, past fault type, fault occurrence time, fault duration, fault influence range; X maint (t) is the maintenance record, periodic maintenance data, replacement component information, cleaning record; Θ is the model parameter, which is obtained by training data optimization.

[0027] As a preferred scheme of the heating surface fault diagnosis and prediction method based on boiler flow field analysis, the flow field simulation results output by the boiler flow field analysis and the heating surface fault diagnosis model are coupled to construct a heating surface fault prediction model, and real-time monitoring data of the heating surface fault prediction of the boiler are output.

[0028] The statistical characteristics of the multivariate time series data coupled by the flow field simulation results output by the boiler flow field analysis and the heating surface fault diagnosis model are extracted as the input of the heating surface fault prediction model, the mean value, variance, maximum value and minimum value of each variable are extracted by the sliding window, the change mode of the data is captured, and the feature vector output will be used as the input of the next step of pattern recognition, and the heating surface fault prediction model constructed adopts the following expression:

[0029]

[0030] Wherein, F i,j is the feature vector of the i th variable in the window j; X i,t is the value of the i th variable at time point t; W is the window size; is the mean value of the i th variable in the window j.

[0031] As a preferred scheme of the boiler flow field analysis-based heating surface fault diagnosis and prediction method, the output fault prediction real-time monitoring data of the boiler heating surface is obtained, and a pattern recognition and fault pattern classification is constructed according to the fault prediction real-time monitoring data, the feature vector output by the heating surface fault prediction model is taken as the input of the pattern recognition, the high-dimensional feature data is reduced in dimension, the main features are extracted, and the calculation complexity is reduced, wherein the dimension reduction is performed through nonlinear principal component analysis, the nonlinear features are extracted, the redundant features are reduced, the main information is retained, and the nonlinear principal component analysis-based pattern recognition is constructed, and the following expression is adopted:

[0032] z = σ (W2·σ (W1·x+b1) +b2)

[0033] wherein z is a nonlinear reduced feature matrix, x is an original feature matrix, W1 and W2 are weight matrices, b1 and b2 are bias vectors, and σ is a nonlinear activation function.

[0034] The reduced feature matrix is taken as the input of the fault pattern classification, the feature data is classified through a Gaussian mixture model, complex fault patterns are recognized, and fault early warning is provided, wherein the Gaussian mixture model is used to classify the feature data and recognize the fault pattern, and the following expression is adopted:

[0035]

[0036] wherein P (X|λ) is the probability that the feature vector X belongs to a certain category, w i is the weight of the i-th Gaussian component, is the probability density function of the i-th Gaussian component, μ i is the mean, and Σ i is the covariance.

[0037] As a preferred scheme of the boiler flow field analysis-based heating surface fault diagnosis and prediction method, the feature information obtained through the flow field analysis is combined with real-time monitoring data to construct an intelligent diagnosis system, potential faults are recognized through the intelligent diagnosis system, and preventive maintenance suggestions are provided; the fault pattern classification result is used for the intelligent diagnosis prediction system to provide reliable fault pattern recognition, wherein a comprehensive fault score is calculated, the fault risk is quantified, the severity of the fault is judged through the comprehensive score, corresponding maintenance measures are provided, the comprehensive fault score is used for the intelligent diagnosis prediction system to determine the range and priority of the maintenance measures, and the fault score formula can adopt the following expression:

[0038]

[0039] wherein S f is a comprehensive fault score, and w iweight of the ith failure mode; C i characteristic score of the ith failure mode; n is the number of failure modes.

[0040] As a preferred scheme of the boiler flow field analysis-based heating surface failure diagnosis and prediction method, wherein: the characteristic information obtained by the flow field analysis is combined with real-time monitoring data to construct an intelligent diagnosis system, potential failures are identified through the intelligent diagnosis system, and preventive maintenance suggestions are provided, the degree of failure risk is determined according to the high and low of the failure score, and different maintenance strategies are adopted,

[0041] Low risk: when S f <the first threshold value, the risk level of the warning level is judged as low risk, the warning information is sent to the relevant departments and the public who may be affected, the low-intensity boiler heating surface failure diagnosis is generated and displayed for risk cognition training;

[0042] Medium risk: when the first threshold value ≤ S f <the second threshold value, the risk level of the warning level is judged as medium risk, the warning information is sent to the relevant departments, the medium-intensity boiler heating surface failure diagnosis is generated and displayed for risk communication and emergency drill, and the emergency response plan is started to prepare evacuation and rescue resources;

[0043] High risk: when the second threshold value ≥ S f , the risk level of the warning level is judged as high risk, the emergency warning is issued to all relevant departments and the public, the high-intensity boiler heating surface failure diagnosis is generated and displayed for evacuation route planning and rescue strategy formulation, and the evacuation program and emergency response measures are immediately started.

[0044] In a second aspect, the embodiments of the present application provide a boiler flow field analysis-based heating surface failure diagnosis and prediction system, which comprises,

[0045] The acquisition module acquires boiler data, analyzes the internal flow field of the boiler, constructs a heating surface failure diagnosis model by capturing flow characteristics and heat transfer characteristics, and outputs historical data and environmental factors of failure occurrence;

[0046] The construction module couples the flow field simulation results output by the boiler flow field analysis and the heating surface failure diagnosis model to construct a heating surface failure prediction model; and outputs real-time monitoring data of failure prediction of the boiler heating surface;

[0047] The prediction module utilizes the characteristic information obtained by the flow field analysis, combines with real-time monitoring data to construct an intelligent diagnosis system, identifies potential failures through the intelligent diagnosis system, and provides preventive maintenance suggestions.

[0048] In a third aspect, an embodiment of the present application provides a computer device, comprising a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the steps of the boiler flow field analysis-based heating surface fault diagnosis and prediction method according to the first aspect of the present application are implemented.

[0049] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, which stores a computer program, and when the computer program is executed by a processor, the steps of the boiler flow field analysis-based heating surface fault diagnosis and prediction method according to the first aspect of the present application are implemented.

[0050] The present application has the following advantages: collecting boiler data and performing flow field analysis: by comprehensively collecting multi-dimensional data such as flow field parameters, heating surface state, and operating parameters, and combining with computational fluid dynamics model to analyze the internal flow field of the boiler, the complex flow and heat transfer process inside the boiler are accurately described. This step provides comprehensive and accurate basic data for subsequent fault diagnosis, significantly improving the reliability and accuracy of fault diagnosis.

[0051] Building a heating surface fault diagnosis model: using flow characteristics and heat transfer characteristics, combining historical fault data and environmental factors, a comprehensive fault diagnosis model is built. This model can capture the dynamic changes of the heating surface state of the boiler, realize early identification of potential faults, and greatly improve the timeliness and accuracy of fault warning.

[0052] Coupling flow field simulation results with fault diagnosis model: by coupling the flow field simulation results with the fault diagnosis model, a more comprehensive and dynamic heating surface fault prediction model is built. This coupling method makes full use of flow field information and historical fault data, significantly improves the accuracy and robustness of the model for fault prediction under complex working conditions.

[0053] Feature extraction and dimensionality reduction: the sliding window method is used to extract the statistical features of time series data, and nonlinear principal component analysis is used for dimensionality reduction, effectively capturing the change pattern of data while reducing the computational complexity. This step significantly improves the computational efficiency of the model, enabling the system to process large amounts of data in real time, making fast fault diagnosis possible.

[0054] Building an intelligent diagnosis system: using Gaussian mixture model to classify feature data, identifying complex fault patterns, and combining with comprehensive fault scoring mechanism to realize quantitative evaluation of fault risk. This method not only improves the accuracy of fault identification, but also provides intuitive and interpretable decision support for operation and maintenance personnel, effectively improving the efficiency and pertinence of fault handling.

[0055] Risk classification and preventive maintenance recommendations: According to the fault score, the risk is divided into low, medium and high three levels, and the corresponding early warning and maintenance strategy is formulated for different risk levels. This grading management method realizes the fine management and control of the boiler operation risk, effectively reduces the equipment failure rate, optimizes the allocation of maintenance resources, and improves the overall operation efficiency.

[0056] In summary, the present application builds a comprehensive, accurate and efficient boiler heating surface fault diagnosis and prediction system through a series of innovative steps such as multi-dimensional data acquisition, flow field analysis, nonlinear modeling, feature extraction and dimension reduction, intelligent diagnosis, etc. The system not only significantly improves the accuracy and timeliness of fault diagnosis, but also realizes the fine management of the boiler operation risk, provides strong technical support for the safe and efficient operation of the boiler, and has important practical value and economic benefits. BRIEF DESCRIPTION OF DRAWINGS

[0057] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0058] Fig. 1 Flow chart of the heating surface fault diagnosis and prediction method based on boiler flow field analysis;

[0059] Fig. 2 Computer device diagram of the heating surface fault diagnosis and prediction method based on boiler flow field analysis. DETAILED DESCRIPTION

[0060] In order to make the above-mentioned purposes, features and advantages of the present application more obvious and easy to understand, the specific embodiments of the present application will be described in detail below with reference to the drawings.

[0061] In the following description, many specific details are set forth in order to provide a thorough understanding of the present application, but the present application can also be implemented in other ways different from the description, and those skilled in the art can make similar generalizations without departing from the connotation of the present application, therefore the present application is not limited to the specific embodiments disclosed below.

[0062] Secondly, the "one embodiment" or "embodiment" referred to herein means that the specific features, structures or characteristics can be included in at least one implementation of the present application. "In one embodiment" appearing in different places in the specification does not mean the same embodiment, nor is it an embodiment that is independent of or mutually exclusive with other embodiments.

[0063] Embodiment 1

[0064] Reference Figs. 1-2 For the first embodiment of the present application, the embodiment provides a heating surface fault diagnosis prediction method based on boiler flow field analysis, comprising,

[0065] S100: Collecting boiler data, analyzing the internal flow field of the boiler, constructing a heating surface fault diagnosis model by capturing flow characteristics and heat transfer characteristics, and outputting historical data and environmental factors of fault occurrence;

[0066] S101: The collection of boiler data includes flow field parameter data, heating surface state data, operating parameter data, environmental parameters, historical fault data, equipment state data, and maintenance records;

[0067] Flow field parameter data: flow velocity distribution, temperature distribution, pressure distribution, flue gas composition, the flue gas composition includes O2, CO2, NO x ;

[0068] Heating surface state data: tube wall temperature, tube wall thickness, surface corrosion degree, and fouling condition;

[0069] Operating parameter data: boiler load, feedwater flow, steam parameter, fuel input quantity and property, the steam parameter includes temperature, pressure, and flow;

[0070] Environmental parameters: ambient temperature, humidity, and atmospheric pressure;

[0071] Historical fault data: past fault type, fault occurrence time, fault duration, and fault influence range;

[0072] Equipment state data: vibration data, noise level, and equipment running time;

[0073] Maintenance records: regular maintenance data, replacement component information, and cleaning records.

[0074] S200: Coupling the flow field simulation results output by the boiler flow field analysis and the heating surface fault diagnosis model to construct a heating surface fault prediction model; outputting real-time monitoring data of the boiler heating surface fault prediction;

[0075] S201: The collection of boiler data, the analysis of the internal flow field of the boiler, the collection of the velocity distribution and the temperature distribution of the fluid in the boiler, the provision of basic data for the analysis of the boiler heating surface, and the construction of the internal flow field analysis of the boiler, which adopts the following expression:

[0076]

[0077] Where, ρ is the density of flue gas, calculated from flue gas composition data; v is the flow velocity distribution, flow field parameter; P is the pressure distribution, flow field parameter; μ is the viscosity of flue gas, calculated from fuel input quantity and property; cp is the specific heat capacity of flue gas, calculated from flue gas composition data; T is the temperature distribution, flow field parameters; k is the thermal conductivity of flue gas, calculated from fuel input and properties; S is the heat source term, calculated from fuel input and boiler load data;

[0078] The flow field and temperature distribution analysis outputs detailed flow characteristics of velocity and temperature distribution, and serves as input for the heat transfer characteristics of the boiler heating surface state and corrosion prediction and the boiler heating surface state and fouling prediction by the change of heating surface tube wall thickness over time, and provides key parameters for building a heating surface fault diagnosis model, and builds a closing thickness change equation combining flow field temperature and chemical reaction, which adopts the following expression:

[0079]

[0080] Wherein, W is the tube wall thickness, heating surface state data; k c is the corrosion constant, calculated from past fault type data; E c is the corrosion activation energy, calculated from past fault type data; R is the gas constant; T is the temperature, the temperature distribution output in the last step; C corrosive is the corrosion substance concentration, flue gas composition in the flow field parameters, flue gas composition including O2, CO2, NO x ; k s is the fouling constant, calculated from past fault type data; C scaling is the fouling substance concentration, flue gas composition in the flow field parameters;

[0081] The heating surface fault diagnosis model is built by capturing flow characteristics and heat transfer characteristics, and outputs historical data and environmental factors of fault occurrence; the coupling results of flow field and temperature distribution and tube wall thickness change are taken as input variables, and the flow field, heating surface state, operating parameters, environmental parameters, historical fault data and maintenance records are used to build a heating surface fault diagnosis model for real-time fault prediction, which adopts the following expression:

[0082]

[0083] Wherein, is the predicted fault state; X flow (t) is the flow field parameter, velocity distribution, temperature distribution, pressure distribution, flue gas composition; X heat (t) is the heating surface state parameter, tube wall temperature, tube wall thickness, surface corrosion degree, fouling condition; X run (t) is the operating parameter, boiler load, feed water flow, steam parameter, fuel input and properties; X env (t) is the environmental parameter, environmental temperature, humidity, atmospheric pressure; X hist(t) is historical failure data, past failure type, failure time, failure duration, failure impact range; X maint (t) is maintenance records, periodic maintenance data, replacement parts information, cleaning records; Θ is model parameters, optimized by training data.

[0084] S:202: The flow field simulation results output by the boiler flow field analysis and the heating surface fault diagnosis model are coupled to construct a heating surface fault prediction model; and fault prediction real-time monitoring data of the boiler heating surface is output;

[0085] The statistical characteristics of the multivariate time series data coupled by the flow field simulation results output by the boiler flow field analysis and the heating surface fault diagnosis model are extracted as the input of the heating surface fault prediction model. The mean, variance, maximum and minimum of each variable are extracted by a sliding window to capture the change pattern of the data. The output feature vector will be used as the input of the next step of pattern recognition. The heating surface fault prediction model is constructed using the following expression:

[0086]

[0087] Where, F i,j is the feature vector of the i-th variable in window j; X i,t is the value of the i-th variable at time point t; W is the window size; is the mean of the i-th variable in window j.

[0088] S203: The fault prediction real-time monitoring data of the boiler heating surface is output. According to the fault prediction real-time monitoring data, pattern recognition and fault pattern classification are constructed. The feature vector output by the heating surface fault prediction model is used as the input of the pattern recognition to reduce the dimension of the high-dimensional feature data, extract the main features, and reduce the calculation complexity. The dimension is reduced by nonlinear principal component analysis to extract nonlinear features, reduce redundant features, retain main information, and construct nonlinear principal component analysis pattern recognition, which uses the following expression:

[0089] z = σ (W2·σ (W1·x+b1) +b2)

[0090] Where, z is the feature matrix after nonlinear dimension reduction; x is the original feature matrix; W1 and W2 are weight matrices; b1 and b2 are bias vectors; σ is a nonlinear activation function;

[0091] The feature matrix after dimension reduction will be used as the input of the fault pattern classification. The feature data is classified by the Gaussian mixture model to identify complex fault patterns and provide fault warning. The feature data is classified by the Gaussian mixture model to identify fault patterns, which uses the following expression:

[0092]

[0093] where P(X|λ) is the probability of the feature vector X belonging to a certain class; w i is the weight of the i-th Gaussian component; is the probability density function of the i-th Gaussian component, with mean μ i and covariance Σ i .

[0094] S300: Using the feature information obtained by flow field analysis, combined with real-time monitoring data to build an intelligent diagnosis system, identify potential faults through the intelligent diagnosis system, and provide preventive maintenance recommendations.

[0095] S301: Using the feature information obtained by flow field analysis, combined with real-time monitoring data to build an intelligent diagnosis system, identify potential faults through the intelligent diagnosis system, and provide preventive maintenance recommendations; through the fault mode classification result, the intelligent diagnosis prediction system is used to provide reliable fault mode identification, wherein the comprehensive fault score is calculated, the fault risk is quantified, the severity of the fault is judged through the comprehensive score, and the corresponding maintenance measures are provided. The comprehensive fault score is used in the intelligent diagnosis prediction system to determine the scope and priority of the maintenance measures, and the fault score formula can be expressed as follows:

[0096]

[0097] where S f is the comprehensive fault score; w i is the weight of the i-th fault mode; C i is the feature score of the i-th fault mode; and n is the number of fault modes.

[0098] S302: Using the feature information obtained by flow field analysis, combined with real-time monitoring data to build an intelligent diagnosis system, identify potential faults through the intelligent diagnosis system, and provide preventive maintenance recommendations, determine the degree of fault risk according to the high and low of the fault score, and take different maintenance strategies,

[0099] Low risk: when S f <first threshold value, the risk level of the warning level is low risk, the warning information is sent to the relevant departments and the public who may be affected, and low-intensity boiler heating surface fault diagnosis is generated and displayed for risk awareness training;

[0100] Medium risk: when the first threshold value ≤ S f <second threshold value, the risk level of the warning level is medium risk, the warning information is sent to the relevant departments, and medium-intensity boiler heating surface fault diagnosis is generated and displayed for risk communication and emergency drill, and the emergency response plan is started to prepare evacuation and rescue resources;

[0101] High risk: When the second threshold ≥ S f , the risk level of the warning level is judged as high risk, an emergency warning is issued to all relevant departments and the public, a high-intensity boiler heating surface fault diagnosis is generated and displayed for evacuation route planning and rescue strategy formulation, and at the same time, an evacuation program and emergency response measures are immediately started.

[0102] Where the following threshold values can be used:

[0103] For the comprehensive fault score S f , we can set the following threshold range:

[0104] Low risk: S f <0.3

[0105] Medium risk: 0.3 ≤ S f <0.7

[0106] High risk: S f ≥0.7

[0107] These threshold settings are based on the following considerations:

[0108] 0.3 as the upper limit of low risk: This value is low, indicating that the system is in good operating condition and only requires routine monitoring and maintenance.

[0109] 0.7 as the lower limit of high risk: This value is high, indicating that the system has already shown significant abnormalities and immediate intervention measures are needed.

[0110] Medium risk interval between 0.3 and 0.7: Provides sufficient warning time for maintenance personnel to take preventive maintenance measures.

[0111] This threshold setting achieves fine-grained classification management of boiler operation risks, helping to optimize maintenance resource allocation and improve equipment reliability. Thresholds can be dynamically adjusted according to actual operating data and expert experience to adapt to the needs of different types of boilers and operating environments.

[0112] Flow field analysis: Refers to the use of computational fluid dynamics (CFD) methods to numerically simulate and analyze the fluid motion, heat transfer and chemical reaction processes inside the boiler.

[0113] Heating surface: Refers to the surface in the boiler that directly contacts high-temperature flue gas and performs heat exchange, such as water wall, superheater, etc.

[0114] Nonlinear principal component analysis: A nonlinear dimensionality reduction technique that can effectively extract nonlinear features from data.

[0115] Multi-source data fusion: By integrating multi-dimensional data such as flow field parameters, heated surface state, and operating parameters, a comprehensive description of the boiler's working state is achieved, providing a rich information foundation for fault diagnosis.

[0116] Flow field-fault model coupling: The flow field simulation results are combined with the fault diagnosis model, making full use of the fluid dynamics information and significantly improving the accuracy of the model in predicting faults under complex working conditions.

[0117] Nonlinear feature extraction: Nonlinear principal component analysis is used for dimension reduction, effectively capturing the nonlinear features in the data and overcoming the limitations of traditional linear methods.

[0118] Intelligent diagnosis system: Gaussian mixture model is used for fault pattern recognition, combined with a comprehensive fault scoring mechanism to accurately quantify fault risks and provide reliable basis for operation and maintenance decisions.

[0119] Through the above innovative methods and key steps, the present invention constructs an intelligent and accurate boiler heating surface fault diagnosis and prediction system, significantly improving the accuracy, timeliness, and interpretability of fault diagnosis, and providing strong technical support for the safe and efficient operation of boilers.

[0120] Further, the present embodiment also provides a heating surface fault diagnosis and prediction system based on boiler flow field analysis, comprising,

[0121] The acquisition module acquires boiler data, analyzes the internal flow field of the boiler, constructs a heating surface fault diagnosis model by capturing flow characteristics and heat transfer characteristics, and outputs historical data and environmental factors of fault occurrence;

[0122] The construction module couples the flow field simulation results output by the boiler flow field analysis and the heating surface fault diagnosis model to construct a heating surface fault prediction model; and outputs real-time monitoring data of fault prediction of the boiler heating surface;

[0123] The prediction module uses the feature information obtained by flow field analysis, combined with real-time monitoring data to construct an intelligent diagnosis system, identifies potential faults through the intelligent diagnosis system, and provides preventive maintenance recommendations.

[0124] The present embodiment also provides a computer device suitable for the heating surface fault diagnosis and prediction method based on boiler flow field analysis, comprising a memory and a processor; the memory is used to store computer executable instructions, and the processor is used to execute the computer executable instructions to realize the heating surface fault diagnosis and prediction method based on boiler flow field analysis as proposed in the above embodiments.

[0125] The computer device can be a terminal, which includes a processor, a memory, a communication interface, a display screen and an input device connected by a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used for wired or wireless communication with external terminals. Wireless communication can be achieved through WIFI, operator network, NFC (near field communication) or other technologies. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen. The input device of the computer device can be a touch layer overlaid on the display screen, or a key, trackball or touchpad arranged on the shell of the computer device. It can also be an external keyboard, touchpad or mouse, etc.

[0126] The embodiment also provides a storage medium having a computer program stored thereon, which, when executed by a processor, implements the method for predicting the fault diagnosis of the heating surface based on the analysis of the flow field of the boiler as proposed in the above embodiment.

[0127] In summary, the boiler data is collected and the flow field is analyzed: by comprehensively collecting multi-dimensional data such as flow field parameters, heating surface state and operating parameters, and combining with computational fluid dynamics model to analyze the internal flow field of the boiler, the complex flow and heat transfer process in the boiler is accurately described. This step provides comprehensive and accurate basic data for subsequent fault diagnosis, significantly improving the reliability and accuracy of fault diagnosis.

[0128] Building a heating surface fault diagnosis model: using flow characteristics and heat transfer characteristics, combining historical fault data and environmental factors, a comprehensive fault diagnosis model is built. The model can capture the dynamic changes of the heating surface state of the boiler, realize the early identification of potential faults, and greatly improve the timeliness and accuracy of fault warning.

[0129] Coupling the flow field simulation results with the fault diagnosis model: by coupling the flow field simulation results with the fault diagnosis model, a more comprehensive and dynamic heating surface fault prediction model is built. This coupling method makes full use of flow field information and historical fault data, significantly improves the accuracy and robustness of the model for fault prediction under complex working conditions.

[0130] Feature extraction and dimensionality reduction: the sliding window method is used to extract the statistical features of time series data, and nonlinear principal component analysis is used for dimensionality reduction, effectively capturing the change pattern of the data while reducing the computational complexity. This step significantly improves the computational efficiency of the model, enabling the system to process large amounts of data in real time, making fast fault diagnosis possible.

[0131] Constructing an intelligent diagnosis system: using Gaussian mixture model to classify feature data, identifying complex fault patterns, and combining comprehensive fault scoring mechanism to realize quantitative evaluation of fault risk. This method not only improves the accuracy of fault identification, but also provides intuitive and interpretable decision support for operation and maintenance personnel, effectively improving the efficiency and pertinence of fault handling.

[0132] Risk classification and preventive maintenance recommendations: according to the level of fault score, the risk is divided into low, medium and high three levels, and the corresponding early warning and maintenance strategy is formulated for different risk levels. This grading management method realizes the fine management and control of boiler operation risk, effectively reduces the equipment failure rate, optimizes the allocation of maintenance resources, and improves the overall operation efficiency.

[0133] In summary, the invention builds a comprehensive, accurate and efficient boiler heating surface fault diagnosis and prediction system through a series of innovative steps such as multi-dimensional data acquisition, flow field analysis, nonlinear modeling, feature extraction and dimension reduction, intelligent diagnosis, etc. This system not only significantly improves the accuracy and timeliness of fault diagnosis, but also realizes the fine management of boiler operation risk, providing strong technical support for the safe and efficient operation of the boiler, and has important practical value and economic benefits.

[0134] Example 2

[0135] Reference Fig. 1 - Fig. 2 As the second embodiment of the invention, this embodiment provides a heating surface fault diagnosis and prediction method based on boiler flow field analysis. In order to verify the beneficial effects of the invention, economic benefit calculation and simulation experiments are carried out for scientific demonstration.

[0136] In order to verify the effectiveness of the heating surface fault diagnosis and prediction method based on boiler flow field analysis proposed by the invention, a 600MW unit boiler of a certain thermal power plant is selected as the research object, and a 6-month experiment is carried out. The boiler is a supercritical pressure once-through boiler with rated evaporation capacity of 1930t / h, main steam pressure of 25.4MPa and main steam temperature of 571℃.

[0137] In the experimental preparation stage, high-precision sensor networks are first installed at key positions of the boiler, including temperature sensors, pressure sensors, flow meters and flue gas analyzers, etc. A total of 150 measuring points are arranged, covering the main heating surface areas such as water wall, superheater and reheater. At the same time, a data acquisition system based on edge computing is built to realize millisecond-level data acquisition and preprocessing.

[0138] During the experiment, the system performed a complete data acquisition and analysis cycle every 5 minutes. First, the sensor data, DCS system data and historical operation data were integrated through multi-source data fusion technology. Then, the improved computational fluid dynamics model was used to analyze the internal flow field of the boiler, which considered complex factors such as multiphase flow, chemical reaction and radiation heat transfer, and the calculation accuracy was improved by about 20% compared with the traditional CFD model.

[0139] Based on the flow field analysis results, a heating surface fault diagnosis model was constructed. The model used a deep learning algorithm, specifically a hybrid structure of long short-term memory network (LSTM) and convolutional neural network (CNN), which could capture both temporal and spatial features. The model training used historical data from the past 3 years, including normal operation data and various fault cases, totaling about 5 million records.

[0140] To improve computational efficiency, a sliding window method was used to extract statistical features of time series data, with a window size of 30 minutes and a step size of 5 minutes. Then, improved kernel principal component analysis (KPCA) was used for nonlinear dimensionality reduction, retaining 95% of the information while reducing the feature dimension from the original 1500 to 50.

[0141] In the fault pattern recognition phase, a deep Gaussian mixture model (DGMM) was used, which significantly improved the recognition ability of complex nonlinear fault patterns by introducing a deep neural network. At the same time, an explainability mechanism based on Shapley value was designed, which could provide intuitive explanations for each diagnosis result.

[0142] Finally, according to the comprehensive fault score S f The risk level was divided into low risk (S f <0.3), medium risk (0.3≤S f <0.7) and high risk (S f ≥0.7). The system automatically generated corresponding warning information and maintenance recommendations according to different risk levels, and presented them to the operation and maintenance personnel through a visual interface.

[0143] To evaluate the performance of the method, it was compared with traditional threshold-based fault diagnosis methods and conventional machine learning methods (support vector machine, SVM). The experimental results are shown in the following table:

[0144] Evaluation index Conventional threshold method SVM method Method of the present invention Fault detection rate (%) 78.5 89.3 96.7 False alarm rate (%) 15.2 8.7 3.2 Early warning time (hours) 2.5 5.8 12.3 Diagnosis accuracy rate (%) 72.3 85.6 94.1 Calculation time (seconds / time) 0.5 3.2 1.8

[0145] Through analysis of the above table data, the following conclusions can be drawn:

[0146] Fault detection rate: The fault detection rate of the method reaches 96.7%, which is 18.2 percentage points higher than the traditional threshold method and 7.4 percentage points higher than the SVM method. This shows that the method can more comprehensively and accurately identify various faults, greatly reducing the occurrence of missed detection.

[0147] False positive rate: The false positive rate of the method is only 3.2%, which is significantly lower than the 15.2% of the traditional threshold method and the 8.7% of the SVM method. This means that the method can effectively reduce unnecessary downtime and maintenance caused by false positives, improving the reliability and economy of the system.

[0148] Early warning time: The method can issue a fault warning 12.3 hours in advance, which is 9.8 hours earlier than the traditional threshold method and 6.5 hours earlier than the SVM method. This provides sufficient time for preventive maintenance for operators, greatly reducing the risk of sudden failure.

[0149] Diagnosis accuracy: The diagnosis accuracy of the method reaches 94.1%, which is 21.8 percentage points higher than the traditional threshold method and 8.5 percentage points higher than the SVM method. This shows that the method not only accurately detects faults, but also accurately judges fault types and locations, providing precise guidance for maintenance work.

[0150] Calculation time: Although the method uses complex flow field analysis and deep learning algorithms, through optimized feature extraction and dimension reduction techniques, the single calculation time is controlled within 1.8 seconds, which is 43.8% faster than the SVM method. This ensures that the system can respond in real time and meet the needs of online monitoring.

[0151] In addition, the method also shows excellent generalization ability and robustness. During the experiment, the system successfully predicted and diagnosed 3 new types of faults that did not appear in the training data, providing valuable decision support for timely handling of these faults. At the same time, through the Shapley value explanation mechanism, the system can provide clear explanations for each diagnosis result, greatly improving the trust and decision-making efficiency of operators on the system.

[0152] In summary, the method is significantly superior to traditional methods and conventional machine learning methods in terms of fault detection rate, false positive rate, early warning time, diagnosis accuracy and other key indicators. By organically combining advanced technologies such as flow field analysis, deep learning, and multi-source data fusion, the method realizes the intelligentization and precision of boiler heating surface fault diagnosis and prediction, providing strong technical support for the safe and efficient operation of boilers, and has important engineering application value.

[0153] It should be noted that the above examples are only used to illustrate the technical solutions of the present application but not limit the present application. Although the present application is described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present application can be modified or equivalently replaced, without departing from the spirit and scope of the technical solutions of the present application, which should be covered in the scope of the claims of the present application.

Claims

1. A method for predicting the failure diagnosis of a heating surface based on the analysis of a boiler flow field, characterized in that: Comprising, Collecting boiler data, analyzing the internal flow field of the boiler, constructing a heating surface fault diagnosis model by capturing flow characteristics and heat transfer characteristics, and outputting historical data and environmental factors of fault occurrence; By collecting the velocity distribution and temperature distribution of the fluid in the boiler, providing basic data for the analysis of the boiler heating surface, and constructing the internal flow field analysis of the boiler, the following expression is used: ； wherein, is the flue gas density, calculated from flue gas composition data; is the flow velocity distribution, a flow field parameter; is the pressure distribution, a flow field parameter; is the flue gas viscosity, calculated from fuel input quantity and properties; is the vector differential operator; is the specific heat capacity of the flue gas, calculated from flue gas composition data; is the temperature distribution, a flow field parameter; is the flue gas thermal conductivity, calculated from fuel input quantity and properties; is the heat source term, calculated from fuel input quantity and boiler load data; The flow field and temperature distribution analysis outputs detailed flow characteristics of velocity and temperature distribution, and serves as the input of the heating surface state change. By the change of the heating surface tube wall thickness with time, the heat transfer characteristics of the boiler heating surface state and corrosion prediction and the boiler heating surface state and fouling prediction are outputted, and the key parameters for constructing the heating surface fault diagnosis model are provided. The tube wall thickness change equation combining flow field temperature and chemical reaction is constructed, which uses the following expression: ； wherein, is the tube wall thickness, the heated surface state data; is the corrosion constant, calculated from past failure type data; is the corrosion activation energy, calculated from past failure type data; is the gas constant; is the temperature, the temperature distribution output from the previous step; is the corrosive substance concentration, the flue gas composition in the flow field parameters, the flue gas composition including O2, CO2, NO x ; is the fouling constant, calculated from past failure type data; is the fouling substance concentration, the flue gas composition in the flow field parameters; The heating surface fault diagnosis model is constructed by coupling the flow field and temperature distribution and the tube wall thickness change as input variables, using the flow field, heating surface state, operating parameters, environmental parameters, historical fault data, and maintenance records. Real-time fault prediction is performed, which uses the following expression: ； wherein, is a predicted failure state; is a flow field parameter, flow rate distribution, temperature distribution, pressure distribution, flue gas composition; is a heating surface state parameter, tube wall temperature, tube wall thickness, surface corrosion degree, fouling condition; is an operation parameter, boiler load, feed water flow, steam parameter, fuel input quantity and property; is an environmental parameter, ambient temperature, humidity, atmospheric pressure; is historical failure data, past failure type, failure occurrence time, failure duration, failure influence range; is maintenance record, periodic maintenance data, replacement component information, cleaning record; is a model parameter, obtained by optimizing training data; The flow field simulation results output by the boiler flow field analysis and the heating surface fault diagnosis model are coupled to construct a heating surface fault prediction model. The real-time monitoring data of the boiler heating surface fault prediction is outputted; By extracting the statistical characteristics of the multivariate time series data coupled by the flow field simulation results output by the boiler flow field analysis and the heating surface fault diagnosis model, as the input of the heating surface fault prediction model, the mean, variance, maximum and minimum statistical characteristics of each variable are extracted by the sliding window. The change pattern of the data is captured, and the feature vector outputted will be used as the input of the next step of pattern recognition. The constructed heating surface fault prediction model uses the following expression: ； in, For the first Several variables in the window The feature vector within; For the first Variables at time points The value; For window size; For the first Several variables in the window The mean within; Using the feature information obtained by flow field analysis, combined with real-time monitoring data, an intelligent diagnosis system is constructed. Through the intelligent diagnosis system, potential faults are identified, and preventive maintenance suggestions are provided.

2. The method for prediction of heating surface failure diagnosis based on boiler flow field analysis according to claim 1, characterized in that: The collected boiler data includes flow field parameter data, heating surface state data, operating parameter data, environmental parameters, historical fault data, equipment state data, and maintenance records; Flow field parameter data: flow velocity distribution, temperature distribution, pressure distribution, flue gas composition, including O2, CO2, NO x ​ Heating surface state data: tube wall temperature, tube wall thickness, surface corrosion degree, and fouling condition; Operating parameter data: boiler load, feedwater flow, steam parameter, fuel input quantity and property, the steam parameter includes temperature, pressure, and flow; Environmental parameters: ambient temperature, humidity, and atmospheric pressure; Historical fault data: past fault type, fault occurrence time, fault duration, and fault impact range; Equipment state data: vibration data, noise level, and equipment running time; Maintenance records: regular maintenance data, replacement component information, and cleaning records.

3. The method for prediction of heating surface failure diagnosis based on boiler flow field analysis according to claim 2, characterized in that: The failure prediction real-time monitoring data of the output boiler heating surface is output. According to the failure prediction real-time monitoring data, a pattern recognition and a failure mode classification are constructed. A feature vector output by the heating surface failure prediction model is taken as an input of the pattern recognition, high-dimensional feature data is reduced in dimension, main features are extracted, and calculation complexity is reduced. Nonlinear principal component analysis is used for dimension reduction, nonlinear features are extracted, redundant features are reduced, main information is retained, and a pattern recognition of the nonlinear principal component analysis is constructed, which adopts the following expression: ； wherein, is a non-linearly reduced feature matrix; is an original feature matrix; and is a weight matrix; and is a bias vector; is a non-linear activation function; The reduced feature matrix is taken as an input of the failure mode classification. A Gaussian mixture model is used for classifying the feature data, identifying a complex failure mode, and providing a failure warning. The Gaussian mixture model is used for classifying the feature data, identifying the failure mode, and adopting the following expression: ； wherein, is a feature vector is a probability that the input data belongs to a certain class; is a weight of the th Gaussian component; is a probability density function of the th Gaussian component with mean and covariance .

4. The method for prediction of heating surface failure diagnosis based on boiler flow field analysis according to claim 3, characterized in that: The feature information obtained by the flow field analysis is combined with the real-time monitoring data to construct an intelligent diagnosis system. Potential failures are identified by the intelligent diagnosis system, and preventive maintenance suggestions are provided. The failure mode classification result is used for an intelligent diagnosis prediction system to provide reliable failure mode identification. A comprehensive failure score is calculated, a failure risk is quantified, the severity of the failure is determined by the comprehensive score, corresponding maintenance measures are provided, the comprehensive failure score is used for the intelligent diagnosis prediction system to determine the range and priority of the maintenance measures, and a failure score formula is constructed, which can adopt the following expression: ； wherein, is a comprehensive failure score; is a weight for the th failure mode; is a characteristic score for the th failure mode; is a number of failure modes.

5. The method for prediction of heating surface failure diagnosis based on boiler flow field analysis as claimed in claim 4, wherein: The feature information obtained by the flow field analysis is combined with the real-time monitoring data to construct an intelligent diagnosis system. Potential failures are identified by the intelligent diagnosis system, and preventive maintenance suggestions are provided. The degree of the failure risk is determined according to the failure score, and different maintenance strategies are adopted. Low risk: when <The first threshold value, the risk level of the warning level is low risk, the warning information is sent to the relevant departments and the public who may be affected, and low intensity boiler heating surface fault diagnosis is generated and displayed for risk awareness training; Medium risk: when the first threshold ≤ < the second threshold, the risk level of the early warning level is judged as medium risk, the early warning information is sent to the relevant departments, the medium intensity boiler heating surface fault diagnosis is generated and displayed for risk communication and emergency drill, and the emergency response plan is started to prepare evacuation and rescue resources; High risk: when the second threshold ≥ When the risk level of the early warning level is judged as high risk, an emergency early warning is issued to all relevant departments and the public, a high-intensity boiler heating surface fault diagnosis is generated and displayed for evacuation route planning and rescue strategy formulation, and an evacuation program and emergency response measures are immediately started.

6. A system for predicting failure diagnosis of a heating surface based on boiler flow field analysis, the system for predicting failure diagnosis of a heating surface based on boiler flow field analysis according to any one of claims 1 to 5, characterized in that: The method further includes the following steps: A collection module collects boiler data, analyzes an internal flow field of the boiler, constructs a heating surface failure diagnosis model by capturing flow characteristics and heat transfer characteristics, and outputs historical data and environmental factors of failure occurrence. A construction module constructs a heating surface failure prediction model by coupling a flow field simulation result output by the boiler flow field analysis and the heating surface failure diagnosis model, and outputs failure prediction real-time monitoring data of the boiler heating surface. A prediction module uses feature information obtained by flow field analysis to construct an intelligent diagnosis system in combination with real-time monitoring data. Potential failures are identified by the intelligent diagnosis system, and preventive maintenance suggestions are provided. 7.A computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the computer device is characterized in that: The processor executes the computer program to realize the steps of the heating surface failure diagnosis prediction method based on the boiler flow field analysis in any one of claims 1 to 5.

8. A computer readable storage medium having stored thereon a computer program, characterized in that: The computer program is executed by the processor to realize the steps of the heating surface failure diagnosis prediction method based on the boiler flow field analysis in any one of claims 1 to 5.

Images