Automatic monitoring system of gas turbine generator set

Abstract

The invention discloses an automatic monitoring system of a gas turbine generator set, and relates to the technical field of turbine generator set monitoring, the system comprises the following components: a microscopic state monitoring unit, a data collection and transmission unit, a data comprehensive analysis unit, a fault source tracing and data analysis unit and a control and display management unit; according to the invention, by using the knowledge graph integrated by the fault source tracing and data analysis unit, chaos theory analysis and cross-domain knowledge migration functions, the fault source can be quickly and accurately traced, including determination of specific component problems and related pre-operation condition changes or indirect influences of other components, and the fault source can be accurately traced through chaos data analysis. According to the method, potential fault risks which are difficult to perceive by a traditional data analysis method are found, in addition, fault diagnosis and maintenance experience knowledge of other similar complex industrial equipment is migrated to the field of gas turbine generator sets by means of a cross-domain knowledge migration technology, and the content of a knowledge graph is further enriched.

Description

Technical Field

[0001] The present invention relates to the technical field of turbine generator set monitoring, and in particular to an automatic monitoring system for a gas turbine generator set. Background Art

[0002] Gas turbine generator sets play an important role in the field of energy production. Their efficient and stable operation is crucial for ensuring energy supply. Key components of such generator sets, such as turbine blades and combustion chambers, operate continuously under complex and harsh working conditions, facing tests of extreme conditions such as high temperature, high pressure, and high-speed rotation. The performance and condition of these components directly affect the operation efficiency and safety of the entire unit. Therefore, real-time monitoring and fault warning of key components of gas turbine generator sets are of great significance.

[0003] The traditional life prediction and fault analysis of gas turbine generator sets mainly rely on the monitoring and analysis of macroscopic operation parameters. However, this method is difficult to accurately reflect the impact of internal microscopic structure changes of components on their life. Microscopic structure evolution, such as crystal structure deformation and microscopic crack initiation, is an important precursor to component performance degradation and fault occurrence, but these changes are often difficult to directly reflect in macroscopic parameters. In addition, the traditional method also has obvious deficiencies in fault analysis and is difficult to quickly and accurately trace the root cause of faults, resulting in the lack of pertinence and efficiency in formulating maintenance strategies. Although there are already some monitoring and analysis means, there is still room for further improvement in comprehensively and deeply excavating various information in monitoring data to accurately predict the root cause of faults and potential risks.

[0004] In view of the above problems, it is necessary to optimize the existing automatic monitoring system for gas turbine generator sets. By installing special microscopic monitoring devices inside or on the surface of key components of gas turbine generator sets, real-time information on the microscopic structure changes of component materials can be obtained, and comprehensive analysis can be carried out in combination with material science theoretical models and big data analysis technologies. Therefore, it is of great significance to develop an automatic monitoring system for gas turbine generator sets that can comprehensively achieve the above characteristics. Summary of the Invention

[0005] The purpose of the present invention is to make up for the deficiencies of the existing technology and provide an automatic monitoring system for a gas turbine generator set. It can accurately predict the remaining service life of components by real-time monitoring the changes in the microscopic structure of key components, and at the same time, use advanced technologies such as knowledge graphs, chaos theory analysis, and cross-domain knowledge transfer to quickly and accurately trace the root cause of faults, discover potential fault risks, and provide a more accurate basis for the maintenance of the unit and component replacement.

[0006] To solve the above technical problems, the present invention provides the following technical solution: An automatic monitoring system for a gas turbine generator set, the system comprising the following components: a microscopic state monitoring unit, a data collection and transmission unit, a data comprehensive analysis unit, a fault root cause tracing and data analysis unit, and a control and display management unit;

[0007] The microscopic state monitoring unit selects an electron microscope probe with high-resolution imaging capabilities and an array composed of various types of micro-nano sensors as microscopic monitoring devices. The sensors include, but are not limited to, strain sensors, capacitance sensors, and resistance sensors. For the key components of the gas turbine generator set, key areas prone to microscopic structure changes are determined based on the component structure, working conditions, and experience, and the monitoring devices are arranged and installed;

[0008] The data collection and transmission unit is equipped with a multi-channel high-precision data collection card, which can interface with the image and electrical signal data output by the microscopic monitoring unit. For image data, a specific collection frame rate is set according to the monitoring accuracy and processing capabilities to ensure the capture of continuous change information of the microscopic structure of key components. For electrical signal data, sampling frequency and resolution collection parameters are set according to the output characteristics of different sensors. By adopting two methods of optical fiber communication and wireless communication, the advantages of each are utilized to enable the data to be quickly transmitted from the collection card to the comprehensive analysis unit under different working conditions;

[0009] The data comprehensive analysis unit includes three modules: image information processing, electrical signal data processing, and life prediction. The image information processing module preprocesses the received images and identifies and quantifies the key features related to the microscopic structure changes of the components based on the algorithm of the convolutional neural network. The electrical signal data processing module restores the electrical signals to parameter values related to the actual physical and chemical property changes according to the sensor output characteristics and calibration models. The life prediction module combines the relevant theoretical models of materials science and big data analysis models, takes the microscopic structure evolution data extracted and restored from the above two modules as input, and predicts the remaining service life of the key components after complex calculations, providing a basis for maintenance decisions;

[0010] The fault root cause tracing and data analysis unit includes a knowledge graph construction and fault root cause tracing subunit, a chaotic data analysis and potential fault risk mining subunit, a chaotic data analysis and potential fault risk mining subunit, and a cross-domain knowledge migration subunit. The knowledge graph construction and fault root cause tracing subunit includes a knowledge graph construction module and a fault root cause tracing module. The knowledge graph construction module collects various materials in the unit field, sorts and classifies them, constructs a knowledge graph according to the internal logical relationship of knowledge, and clarifies various association relationships between entities and relationships. The fault root cause tracing module starts from the fault warning data, uses the knowledge graph for reverse reasoning, determines relevant entities and relationships through intelligent algorithms, and traces back reversely to identify the cause of the fault to determine the fault root cause. The chaotic data analysis and potential fault risk mining subunit includes a chaotic feature parameter calculation module and an operating state judgment and potential fault risk mining module. The chaotic feature parameter calculation module determines parameters for the received data and calculates chaotic feature parameters using corresponding algorithms. The operating state judgment and potential fault risk mining module judges the operating state of the unit based on these parameters, mines potential fault risks, and comprehensively judges the possibility and severity of the risks in combination with other monitoring data. The cross-domain knowledge migration subunit includes a knowledge sorting and extraction module, a mapping relationship establishment module, and a knowledge fusion module. The knowledge sorting and extraction module collects materials in other industrial equipment fields and organizes and extracts knowledge on fault modes, operating parameter changes, and maintenance measures. The mapping relationship establishment module compares the knowledge content of the two fields, calculates the correlation degree, and establishes an accurate mapping relationship. The knowledge fusion module integrates the knowledge of the source field into the knowledge graph of the gas turbine generator set field based on this relationship, thereby improving the accuracy of fault root cause tracing and potential risk mining;

[0011] The control and display management unit includes a control instruction generation module and an information display module. The control instruction generation module sets a threshold interval based on the prediction result of the remaining service life of the key components by the data comprehensive analysis unit and the fault root cause information traced by the fault root cause tracing and data analysis unit. The information display module intuitively displays information on the microscopic structure evolution of key components, various monitoring parameters, predicted remaining service life, fault root cause tracing results, and chaotic data analysis results through a display device.

[0012] Further, in the data collection and transmission unit, for image data, a specific collection frame rate is set according to the monitoring accuracy and processing capacity, and the image data frame rate optimization formula is used to determine the frame rate, where V c is the characteristic speed of the microscopic structure change of the key component, L s is the characteristic length scale of the microscopic structure change of the key component, D v represents the effective utilization rate of the data transmission bandwidth, and T p is the available storage time period of the data storage device considering the capacity of the data storage device and the data storage strategy.

[0013] Further, the data collection and transmission unit uses two methods of optical fiber communication and wireless communication. In terms of optical fiber communication, optical fiber communication is selected as the main transmission means. In terms of wireless communication, for special occasions where wiring is inconvenient, such as rotating turbine blades, a communication method based on ZigBee or Wi-Fi is adopted. By setting up wireless transmission nodes near the components, the collected data is wirelessly transmitted to the data comprehensive analysis unit.

[0014] Further, in the data comprehensive analysis unit, the image information processing module preprocesses the received images and identifies and quantifies the key features related to the microstructural changes of the components based on the algorithm of the convolutional neural network. The algorithm formula is: where, Q f represents the quantization value of a certain microstructural feature in the image, w i is the weight of the i-th neuron related to this feature in the CNN model, A i is the activation value of the i-th neuron, c is the bias term, and N represents the total number of neurons related to this microstructural feature.

[0015] Further, in the data comprehensive analysis unit, the life prediction module combines the relevant theoretical models of materials science and the big data analysis model, takes the microstructural evolution data extracted and restored from the above two modules as the input, and predicts the remaining service life of the key components after complex calculations. The calculation formula is: where, L r represents the remaining service life of the key component, L 0 is the initial designed service life of the key component, α i is the weight coefficient of the i-th characteristic factor related to the microstructural evolution, F i is the actual value of the i-th characteristic factor related to the microstructural evolution at the current moment, β j is the weight coefficient of the j-th characteristic factor related to the macroscopic operating parameters, M j is the actual value of the j-th characteristic factor related to the macroscopic operating parameters at the current moment, γ k is the weight coefficient of the k-th characteristic factor related to other influencing factors, C k is the actual value of the k-th characteristic factor related to other influencing factors at the current moment, I represents the total number of characteristic factors related to the microstructural evolution, J represents the total number of characteristic factors related to the macroscopic operating parameters, K represents the total number of characteristic factors related to other influencing factors, and through comprehensive analysis and calculation, the remaining service life of the key component is predicted.

[0016] Furthermore, in the fault root cause tracing and data analysis unit, the fault root cause tracing module starts from the fault warning data, uses the knowledge graph for reverse reasoning, determines relevant entities and relationships through intelligent algorithms and traces back reversely to identify the cause of the fault. Its algorithm formula is: where P(F root |E 1 ,E 2 ,…,E n ) represents the conditional probability that the root cause of the fault is F 1 ,E 2 ,…,E n occurring under the condition that a series of associated events E root have occurred. P(E i |F root ) is the conditional probability that the associated event E τoot occurs under the condition that the root cause of the fault is F i . P(F root is the prior probability of the occurrence of the root cause of the fault F root itself. P(E i |F j ) is the conditional probability that the associated event E j occurs under the condition that the root cause of the fault is F i . P(F j ) is the prior probability of the occurrence of other root causes of the fault F j itself. n represents the number of associated events related to the current fault manifestation, and m represents the number of all possible root causes of the fault.

[0017] Furthermore, in the fault root cause tracing and data analysis unit, the chaotic characteristic parameter calculation module determines parameters for the received data and calculates the chaotic characteristic parameters using corresponding algorithms. Its algorithm formula is: where C state represents the chaos degree index of the unit operation state. λ max is the maximum value of the calculated Lyapunov exponents. λ i is the i-th Lyapunov exponent, used to comprehensively consider the changes in the unit operation state in all directions. D c is the currently calculated correlation dimension. D c reflects the current complexity of the system under the unit operation state. D min is the minimum value of the correlation dimension under the normal operation state. When D c is less than D min , it indicates that the unit operation state is abnormal. D max is the maximum value of the correlation dimension in the chaotic state. Similarly, by analyzing a large number of gas turbine generator sets that may exhibit chaos, when Dc Less than D min When it is, it indicates that the operating state of the unit is in a chaotic state, and N represents the number of Lyapunov exponents calculated.

[0018] Furthermore, in the fault root cause tracing and data analysis unit, the mapping relationship establishment module and the knowledge fusion module compare the knowledge contents of the two fields, calculate the correlation degree and establish an accurate mapping relationship. The correlation degree calculation formula is: Among them, R ij represents the correlation degree between the i-th knowledge element in the source field knowledge and the j-th knowledge element in the gas turbine generator set field knowledge, w k is the weight coefficient of the k-th characteristic dimension, S ik is the eigenvalue of the i-th knowledge element in the source field knowledge on the k-th characteristic dimension, S jk is the eigenvalue of the j-th knowledge element in the gas turbine generator set field knowledge on the k-th characteristic dimension, and K represents the total number of characteristic dimensions used to measure the knowledge correlation degree.

[0019] Compared with the prior art, the gas turbine generator set automatic monitoring system has the following beneficial effects:

[0020] First, by using the knowledge graph, chaotic theory analysis, and cross-domain knowledge transfer functions integrated in the fault root cause tracing and data analysis unit, the present invention can quickly and accurately trace the fault root cause, including determining specific component problems and related pre-operation condition changes or indirect influences of other components, etc. At the same time, through chaotic data analysis, potential fault risks that are difficult to detect by traditional data analysis methods are discovered. In addition, with the help of cross-domain knowledge transfer technology, the fault diagnosis and maintenance experience knowledge of other similar complex industrial equipment is transferred to the gas turbine generator set field, further enriching the content of the knowledge graph and improving the accuracy and efficiency of fault root cause tracing and potential risk mining, providing a more comprehensive means for the fault diagnosis and maintenance of gas turbine generator sets and improving the reliability and safety of the unit.

[0021] Second, by installing microscopic monitoring devices inside or on the surface of key components of the gas turbine generator set, the present invention can monitor the microscopic structure changes of component materials in real time and accurately, such as crystal structure deformation, microscopic crack initiation, etc. Combining with the material science theory model and big data analysis technology, the remaining service life of key components can be accurately predicted, not only making up for the deficiencies of traditional life prediction methods based on macroscopic operating parameters, but also being able to discover potential fault risks in advance and providing sufficient maintenance time for operation and maintenance personnel.

[0022] Other advantages, objects, and features of the present invention will be set forth in part in the following description, and in part will be obvious to those skilled in the art from a study of the following, or may be learned by practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are only some embodiments of the present invention, and for those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0024] Figure 1 It is a schematic diagram of an automatic monitoring system for a gas turbine generator set;

[0025] Figure 2 It is a flow operation diagram of an automatic monitoring system for a gas turbine generator set. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0026] The following will clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, rather than all embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts fall within the scope of protection of the present invention.

[0027] Embodiment 1

[0028] This embodiment details the specific application of an automatic monitoring system for a gas turbine generator set in the maintenance and fault troubleshooting of turbine blades. By adopting the present invention, the monitoring and maintenance level of the turbine blade state is significantly improved, the risk of faults is reduced, and the stable operation of the unit is ensured.

[0029] For the key component of the turbine blade, the microscopic state monitoring unit selects a high-resolution electron microscope probe and a micro-nano sensor array for microscopic monitoring. The electron microscope probe is installed at key parts such as the blade root, blade tip, and leading edge of the turbine blade, and is fixed by a specially designed mounting fixture to ensure that the structural details at the microscopic level of the blade material, such as crystal structure deformation, can be clearly presented. The micro-nano sensor array is evenly embedded in the blade surface and is tightly combined with the blade material by a special encapsulation process, and can sense in real time the changes in physical and chemical properties such as dislocation density changes and microscopic crack initiation that occur in the component material during operation.

[0030] The data collection and transmission unit is equipped with a high-precision data collection card, which is connected to the image data output by the electron microscope probe and the electrical signal data output by the micro-nano sensor array at the same time. For the image data, according to the accuracy requirements for monitoring the microscopic structure changes of the turbine blade and the data processing ability, the image data frame rate optimization formula is used to set the collection frame rate. By analyzing and statistically processing a large amount of historical monitoring data, the characteristic speed V c (such as calculated by statistically analyzing the crystal structure deformation rate, microscopic crack propagation speed, etc. in the past), characteristic length scale L s (such as statistically calculating the average length of microscopic cracks and determining the average size of the deformation area for crystal structure deformation) of the microscopic structure changes of the turbine blade are determined. At the same time, according to the actual situation of the current data transmission system, the effective utilization rate D v (considering factors such as the selected communication method, the loss of the transmission line, and the interference during data transmission) of the data transmission bandwidth and the available storage time period T p (determined based on the capacity of the data storage device and the current data storage strategy) of the data storage device are calculated. These values are substituted into the formula to calculate an appropriate collection frame rate to ensure that the continuous changes of the microscopic structure of the turbine blade at different times can be completely captured. For the electrical signal data, according to the output characteristics of different types of micro-nano sensors, the sampling frequency is set to 1000 times per second and the resolution is 16 bits to accurately collect the changes of relevant parameters such as dislocation density. Fiber optic communication is used as the wired communication method to transmit the collected data to the data comprehensive analysis unit. A fiber optic interface is set near the gas turbine generator set, and the signal output by the collection card is transmitted to the data comprehensive analysis unit through the fiber optic cable.

[0031] In the data comprehensive analysis unit, after the image information processing module receives the image data, it first normalizes the image to make the image pixel values within a specific standard range, then performs denoising processing to improve the image quality, and then uses an image recognition algorithm based on convolutional neural network (CNN) to perform feature extraction and analysis on the preprocessed image. Through the trained CNN model, key features related to the microscopic structure changes of the blade, such as crystal structure deformation and microscopic cracks, are identified, and according to the microscopic structure feature quantization formula based on convolutional neural network (CNN) these features are quantized. Among them, w i is the weight of the i-th neuron related to this feature in the CNN model, which is obtained through training and learning on a large number of sample images marked with known microscopic structure features. During actual operation, as the microscopic structure changes of the turbine blade continue to change, new marked sample images are regularly collected to retrain the model and update w iand the value of c to ensure that the model can accurately quantify newly emerged microstructural features and accurately reflect the changes in the blade microstructure, A i is the activation value of the i-th neuron, which is calculated in real time during the operation of the model. c is the bias term, a fixed parameter learned by the CNN model during training. The electrical signal data processing module performs parsing and restoration operations on the received electrical signal data according to the output characteristics of different types of sensors and the pre-established precise calibration model, and accurately restores the electrical signal to actual parameter values directly related to the changes in the physical and chemical properties of the blade material, such as the actual dislocation density, microcrack initiation rate, etc. The life prediction module combines relevant theoretical models in the field of materials science such as crystal plasticity theory and fracture mechanics theory, as well as a big data analysis model pre-established by collecting, organizing a large amount of data and training using machine learning algorithms, and takes the microstructural evolution data extracted and restored from the image information processing module and the electrical signal data processing module as input, and predicts the remaining service life of the turbine blade according to the remaining service life prediction formula of the component based on multiple factors to predict the remaining service life of the turbine blade, where L r represents the remaining service life of the turbine blade, L 0 is the initial design service life of the turbine blade, α i is the weight coefficient of the i-th characteristic factor related to microstructural evolution, obtained by training and learning from a large amount of historical operation data and the corresponding actual service life data of the blade, F i is the actual value of the i-th characteristic factor related to microstructural evolution at the current moment (such as the specific value of the dislocation density at a certain moment, the length of the microcrack, etc.), β j is the weight coefficient of the j-th characteristic factor related to macroscopic operation parameters, obtained by training and learning from historical data, M j is the actual value of the j-th characteristic factor related to macroscopic operation parameters at the current moment (such as the actual operating temperature and pressure values on a certain day, etc.), γ k is the weight coefficient of the k-th characteristic factor related to other influencing factors, obtained by training and learning from historical data, C k is the actual value of the k-th characteristic factor related to other influencing factors at the current moment (such as the current humidity value corresponding to environmental factors, the current number of repairs corresponding to the maintenance history, etc.). During operation, as changes occur in aspects such as environmental factors, operating conditions, and maintenance measures, collect the latest data regularly, retrain the machine learning algorithm, and update each weight coefficient and actual value to ensure that the prediction formula can accurately predict the remaining service life of the turbine blade.

[0032] In the fault root cause tracing and data analysis unit, the knowledge graph construction module of the knowledge graph construction and fault root cause tracing sub-unit comprehensively collects various relevant materials about turbine blades in the field of gas turbine generator sets, including design drawings, technical manuals, operation logs, maintenance records, etc. It systematically sorts and classifies the information in these materials, comprehensively integrates various types of knowledge such as components, operation parameters, fault types, and maintenance measures related to turbine blades, and represents them in a structured form of entities and relationships. Using professional knowledge graph construction tools (such as Neo4j, etc.), according to the internal logical relationships between knowledge, different types of knowledge are accurately mapped into entities and relationships in the knowledge graph, constructing a complete, accurate, and richly relevant knowledge graph, which clarifies various association relationships between the faults of turbine blades and changes in operation parameters that may cause the faults, states of other related components, etc. At the same time, the cross-domain knowledge migration sub-unit uses cross-domain knowledge migration technology to migrate the relevant knowledge of turbine blades in the field of aero-engines to the field of gas turbine generator sets. By comparing the similarities of turbine blades, the correlations of operation parameters, and the similarities of fault modes in the two fields, mapping relationships are established, and according to the correlation degree calculation formula of cross-domain knowledge migration Measure the association degree between different knowledge. Among them, R ij represents the association degree between the i-th knowledge element in the knowledge of the aero-engine field and the j-th knowledge element in the knowledge of the gas turbine generator set field. w k is the weight coefficient of the k-th feature dimension, obtained by analyzing and statistically processing a large number of relevant materials and actual application cases. S ik is the eigenvalue of the i-th knowledge element in the knowledge of the aero-engine field on the k-th feature dimension. S j k is the eigenvalue of the j-th knowledge element in the knowledge of the gas turbine generator set field on the k-th feature dimension. K represents the total number of feature dimensions used to measure the knowledge association degree. In the actual application process, with the further understanding and application of knowledge in different fields, and the changes in the operation characteristics of gas turbine generator sets themselves, a large number of relevant materials and actual application cases are regularly re-analyzed to re-determine the weight coefficients w k of different feature dimensions, and at the same time, re-determine the eigenvalues S ik 、S jk, to ensure that the correlation between the knowledge elements of the two fields can be accurately calculated. The fault root cause tracing module in the knowledge graph construction and fault root cause tracing subunit receives the fault warning related data from the data comprehensive analysis unit, takes the current fault manifestation as the tracing starting point, and uses the constructed knowledge graph for reverse reasoning analysis. Through intelligent algorithms, the entities and relationships related to the current fault manifestation are quickly determined, and then traced back along these relationships, and the various related factors related to the current fault manifestation are comprehensively analyzed. In the tracing process, according to the logical relationship between the entities and relationships in the knowledge graph, the possible causes of the fault are gradually checked, and the various causes that may cause the fault are gradually checked. Finally, the specific components with problems are accurately determined (such as determining that the fault is caused by micro cracks in a certain area on the turbine blade), and clarifying which pre-operating conditions have changed or other components have indirectly affected the fault, so as to achieve rapid and accurate tracing of the root cause of the fault. In the tracing process, according to the fault root cause tracing probability calculation formula based on the knowledge graph, Quantify the relationship between the root cause of the fault and various related factors, where P(F root |E 1 ,E 2 ,…;E n ) means that in a known series of related events E 1 ,E 2 ,…,E n In the event of a fault, the root cause is F root The conditional probability, P(E i |F root ) is the fault source is F root Under the condition of i The conditional probability of occurrence is obtained by statistically analyzing a large amount of gas turbine generator set operation data, maintenance records and other information. root ) is the root cause of the fault F root The prior probability of its occurrence is obtained by statistical analysis of historical data, P(E i |F j ) is the fault source is F j (j=1,2,…,m, indicating other possible fault sources), the associated event E i The conditional probability of occurrence is obtained by statistical analysis of historical data, P(F j ) are other possible sources of failure F jThe prior probability of its occurrence is obtained by statistically analyzing historical data. n represents the number of associated events related to the current fault manifestation, and m represents the number of all possible fault sources. As the gas turbine generator unit operates, new fault types may emerge, or the relationship between the original fault sources and associated events may change. Regularly collect the latest operation data, maintenance records, etc., and then conduct statistical analysis again to update the prior probability P(F root )、P(F j ) and the conditional probability P(E i |F root )、P(E i |F j) to ensure that the likelihood of different fault sources can be accurately calculated under the occurrence of known associated events. In the chaotic data analysis and potential fault risk mining subunit, the chaotic feature parameter calculation module deeply analyzes the data transmitted from the data comprehensive analysis unit using chaotic theory. For time series data, according to the distribution law, periodicity, and other characteristics of the data, key parameters such as appropriate delay time and embedding dimension are scientifically determined. Then, using corresponding mathematical algorithms, the monitoring data is reconstructed in phase space with the determined delay time and embedding dimension, and then chaotic feature parameters such as Lyapunov exponents and correlation dimensions are accurately calculated. During the calculation process, through multiple verifications and parameter adjustments, it is ensured that the calculated chaotic feature parameters can truly reflect the chaotic characteristics of the unit operation state. For example, by deeply analyzing a large amount of normal operation data of gas turbine generator sets, the reference values of Lyapunov exponents and correlation dimensions are initially determined. The monitoring data during the normal operation of the unit is collected, and the Lyapunov exponents and correlation dimensions are calculated using relevant algorithms of chaotic theory. Then, statistical analysis is performed on these Lyapunov exponents and correlation dimensions to find characteristic values such as the maximum and minimum values as the reference values of the initially determined Lyapunov exponents and correlation dimensions. During the actual operation process, as the operating conditions of the gas turbine generator set change, such as load changes and ambient temperature changes, the latest monitoring data is regularly re-collected, the Lyapunov exponents and correlation dimensions are recalculated, and their reference values are adjusted according to the new calculation results to ensure that the chaotic characteristics of the unit operation state can be accurately reflected. The operation state judgment and potential fault risk mining module judges whether the unit operation state is at the edge of chaos or has exhibited chaotic phenomena based on the calculated chaotic feature parameters. Specifically, when the Lyapunov exponent is greater than zero or the correlation dimension exhibits a specific change pattern (such as exceeding the normal range, showing abnormal fluctuations, etc.), it indicates that the unit operation state may be at the edge of chaos or has exhibited chaotic phenomena. At this time, further explore the hidden non-linear and non-periodic complex laws. Through in-depth analysis and mining of the data, potential fault risks that are difficult to detect by traditional data analysis methods are discovered. During the mining process, professional data analysis tools and algorithms are used to comprehensively analyze the monitoring data, and combined with chaotic feature parameters and other relevant monitoring data, the likelihood and severity of potential fault risks are comprehensively judged.

[0033] In the control and display management unit, the control instruction generation module sets multiple different threshold intervals based on key information such as the remaining service life of the turbine blade predicted by the data comprehensive analysis unit and the fault source traced by the fault source tracing and data analysis unit. For example, for the key component of the turbine blade, considering its importance and the significant impact that a fault may have on the unit operation, the maintenance reminder threshold T mSet to about 30% of the remaining service life. When the remaining service life of the turbine blade is lower than T m , an automatic maintenance reminder instruction is generated to notify the operation and maintenance personnel to inspect and maintain the turbine blade. If after tracing the root cause of the fault and the conditional probability P(F root |E 1 ,E 2 ,…,E n ) calculated according to the fault root cause tracing probability calculation formula based on the knowledge graph is greater than the set fault root cause confirmation threshold T p (such as set to about 0.8), a component replacement suggestion instruction is generated to guide the operation and maintenance personnel to replace the relevant components in time to avoid failures of the gas turbine generator set due to component damage. At the same time, combined with the fault root cause tracing results, if it is determined that a certain operating parameter abnormality causes the fault, an instruction to adjust the operating parameter can also be generated to guide the operation and maintenance personnel to perform corresponding parameter adjustment operations to restore the normal operating state of the unit. During the operation of the gas turbine generator set, as the performance and fault conditions of the turbine blade are further understood and the operating conditions change, regularly re-evaluate the wear and fault risks of the components and adjust the values of T m and T p to ensure the rationality of the control instruction generation. The information display module intuitively displays the microscopic structure evolution of the turbine blade through the liquid crystal display screen, including presenting visual information such as crystal structure deformation images and microscopic crack development conditions, and at the same time displays various monitoring parameters, such as specific values of dislocation density and microscopic crack initiation rate, and displays the predicted remaining service life and important information such as the fault root cause tracing results and chaotic data analysis results obtained by the fault root cause tracing and data analysis unit, so that the operation and maintenance personnel can comprehensively and clearly understand the current state of the turbine blade and the specific conditions and potential risks related to the fault, thereby providing an intuitive reference basis for the operation and maintenance personnel to take corresponding measures.

[0034] In summary, the present invention plays an important role in the monitoring, maintenance, fault troubleshooting and potential risk mining of the turbine blade of the gas turbine generator set, effectively improves the control ability of the turbine blade state, and improves the reliability and safety of the operation of the gas turbine generator set, reducing the fault loss.

[0035] Embodiment 2

[0036] This embodiment details the application of an automatic monitoring system of a gas turbine generator set in the maintenance and fault analysis of the combustion chamber of the gas turbine generator set. Through the present invention, the monitoring accuracy of the combustion chamber state is effectively improved, the root cause of the fault is accurately traced, and potential fault risks are mined, providing strong support for the stable operation of the gas turbine generator set.

[0037] The microscopic state monitoring unit selects an electron microscope probe and a micro-nano sensor array for microscopic monitoring of the combustion chamber, a key component. After pretreatment (such as polishing, cleaning, etc.) on the inner wall surface of the combustion chamber, a micro-nano sensor array composed of strain sensors, capacitance sensors, etc. is embedded in a grid layout. A special encapsulation process is used to tightly bond the sensors to the combustion chamber wall, ensuring that the sensors can accurately sense the physical and chemical property changes occurring in the combustion chamber wall during operation, such as changes in dislocation density and the initiation of microscopic cracks. At the same time, an electron microscope probe is installed at a suitable position in the combustion chamber to clearly observe the crystal structure deformation of the combustion chamber wall. The information obtained by these monitoring devices is crucial for understanding the microscopic state of the combustion chamber and subsequent analysis and maintenance decisions.

[0038] The data collection and transmission unit is equipped with a high-precision data collection card for collecting the image data output by the electron microscope probe and the electrical signal data output by the micro-nano sensor array. For the image data, a suitable collection frame rate is set according to the monitoring accuracy requirements for the microscopic structure changes in the combustion chamber and the data processing ability to ensure that the continuous changes in the microscopic structure of the combustion chamber at different times can be completely captured. For the electrical signal data, according to the output characteristics of different types of micro-nano sensors, such as signal amplitude range, frequency characteristics, sensitivity, etc., the corresponding collection parameters are accurately set. For example, the sampling frequency is set to 800 times per second and the resolution is 12 bits to accurately collect the subtle changes in relevant parameters such as dislocation density changes and the initiation of microscopic cracks, so that the collected data can truly reflect the actual state of the combustion chamber component materials. In terms of data transmission, optical fiber communication is used as the main method of wired communication. An optical fiber interface is set near the gas turbine generator set, and the signal output by the collection card is transmitted to the data comprehensive analysis unit through the optical fiber, making full use of the advantages of optical fiber communication, such as high bandwidth, low loss, and strong anti-interference ability, to ensure the integrity and accuracy of data transmission. At the same time, considering the areas in the combustion chamber where wiring is inconvenient, for these special situations, a wireless communication method based on the ZigBee wireless communication protocol is adopted, and wireless transmission nodes are reasonably set in the corresponding areas to wirelessly transmit the collected data to the data comprehensive analysis unit, ensuring the timeliness and reliability of data transmission under complex working conditions.

[0039] In the data comprehensive analysis unit, after the image information processing module receives the image data from the data collection and transmission unit, it first normalizes the image to make the image pixel values within a specific standard range for convenient subsequent processing. Then, it performs denoising processing to remove the noise interference in the image through a filtering algorithm and improve the image quality. Next, it applies advanced image recognition technology, specifically the image recognition algorithm based on convolutional neural network (CNN), to extract and analyze the features of the preprocessed image. Through the trained CNN model, it accurately identifies the key features related to the microscopic structure changes of the combustion chamber, such as crystal structure deformation, microscopic cracks, etc. Moreover, it quantifies these identified features and converts them into specific parameter values that can be used for subsequent analysis and calculation. For example, it represents features such as the length of microscopic cracks and the degree of crystal structure deformation in specific numerical forms. During the actual operation process, as the microscopic structure changes of the combustion chamber continue to vary, new labeled sample images need to be collected regularly to retrain the CNN model to ensure that the model can accurately quantify the newly emerging microscopic structure features and accurately reflect the changes in the microscopic structure of the combustion chamber. For the received electrical signal data, the electrical signal data processing module performs parsing and restoration operations based on the output characteristics of different types of sensors and the pre-established precise calibration model. Through this process, the electrical signal can be accurately restored to actual parameter values directly related to the changes in the physical and chemical properties of the combustion chamber components, such as dislocation density and microscopic crack initiation rate, for comprehensive analysis with other monitoring data. The remaining life prediction module combines relevant theoretical models in the field of materials science, such as crystal plasticity theory and fracture mechanics theory, and the big data analysis model pre-trained through a large amount of data collection, collation, and machine learning algorithms (such as support vector machine (SVM), random forest (RF), etc.). Taking the microscopic structure evolution data extracted and restored from the image information processing module and the electrical signal data processing module as input, after a series of comprehensive analysis and calculations, it predicts the remaining service life of the combustion chamber. During this process, as changes occur in aspects such as environmental factors, operating conditions, and maintenance measures, new data needs to be collected regularly to retrain the machine learning algorithm and update relevant parameters such as weight coefficients and actual values to ensure that the prediction results can accurately reflect the actual remaining service life of the combustion chamber and provide a reliable basis for subsequent maintenance decisions.

[0040] In the fault root cause tracing and data analysis unit, the knowledge graph construction module of the knowledge graph construction and fault root cause tracing sub-unit comprehensively collects various relevant materials about the combustion chamber in the field of gas turbine generator sets, including design drawings, technical manuals, operation logs, maintenance records, etc. It systematically sorts and classifies the information in these materials, comprehensively integrates various types of knowledge such as the components, operation parameters, fault types, and maintenance measures related to the combustion chamber, and then represents it in a structured form of entities and relationships. Using professional knowledge graph construction tools (such as Neo4j, etc.), according to the internal logical relationships between the knowledge, different types of knowledge are accurately mapped into entities and relationships in the knowledge graph, and a complete, accurate and richly relevant knowledge graph is constructed. In this knowledge graph, various correlation relationships between the faults of the combustion chamber and the changes in operation parameters that may cause the faults, the states of other relevant components, etc. are clarified. At the same time, using cross-domain knowledge transfer technology, the knowledge related to the combustion chamber in the field of aero-engines is transferred to the field of gas turbine generator sets. By comparing the similarities of the combustion chambers, the correlations of operation parameters, and the similarities of fault modes in the two fields, an accurate and reasonable mapping relationship is established. When carrying out knowledge transfer, the mature experience and knowledge about the combustion chamber in the aero-engine field can be used to enrich the content of the knowledge graph in the field of gas turbine generator sets and improve the accuracy and efficiency of fault root cause tracing. After receiving the fault warning-related data transmitted from the data comprehensive analysis unit, the fault root cause tracing module uses the current fault manifestation as the tracing starting point and conducts reverse reasoning analysis using the constructed knowledge graph. First, through intelligent algorithms, it quickly determines the entities and relationships related to the current fault manifestation. These algorithms can quickly search and match the elements related to the current fault manifestation in the knowledge graph, and then trace back along these relationships to comprehensively analyze various correlation factors related to the current fault manifestation. During the tracing process, according to the logical relationships between the entities and relationships in the knowledge graph, various possible causes of the fault are gradually investigated, and finally the specific component with problems is accurately determined (for example, it is determined that a certain area on the combustion chamber wall has microscopic cracks leading to the fault), and it is clarified which pre-operation condition changes or indirect influences of other components cause the fault to occur, realizing the rapid and accurate tracing of the fault root cause. The chaotic feature parameter calculation module in the chaotic data analysis and potential fault risk mining sub-unit conducts in-depth analysis on the data transmitted from the data comprehensive analysis unit using chaos theory. For time series data, according to the distribution law, periodicity and other characteristics of the data, key parameters such as the appropriate delay time and embedding dimension are scientifically determined, and then corresponding mathematical algorithms are used to perform phase space reconstruction operations on the monitoring data using the determined delay time and embedding dimension, and then the chaotic feature parameters such as Lyapunov exponents and correlation dimensions are accurately calculated. During the calculation process, through multiple verifications and parameter adjustments, it is ensured that the calculated chaotic feature parameters can truly reflect the chaotic characteristics of the unit operation state. For example,The reference values of Lyapunov exponents and correlation dimensions are initially determined by deeply analyzing a large amount of normal operation data of gas turbine generator sets. During the normal operation of the unit, the monitoring data is collected, and the Lyapunov exponents and correlation dimensions are calculated using relevant algorithms of chaos theory. Then, statistical analysis is carried out on these Lyapunov exponents and correlation dimensions to find out the characteristic values such as the maximum and minimum values, which are used as the reference values of the initially determined Lyapunov exponents and correlation dimensions. During the actual operation process, with the change of the operating conditions of the gas turbine generator set, such as load change, ambient temperature change, etc., it is necessary to regularly collect the latest monitoring data, recalculate the Lyapunov exponents and correlation dimensions, and adjust their reference values according to the new calculation results to ensure that the chaotic characteristics of the unit operation state can be accurately reflected. The operation state judgment and potential fault risk mining module judges whether the operation state of the unit is at the edge of chaos or chaotic phenomenon has occurred according to the calculated chaotic characteristic parameters. Specifically, when the Lyapunov exponent is greater than zero or the correlation dimension shows a specific change rule (such as exceeding the normal range, abnormal fluctuation, etc.), it indicates that the operation state of the unit may be at the edge of chaos or chaotic phenomenon has occurred. At this time, further explore the hidden non-linear and non-periodic complex rules. Through in-depth analysis and mining of the data, potential fault risks that are difficult to detect by traditional data analysis methods are discovered. During the mining process, professional data analysis tools and algorithms should be used to comprehensively analyze the monitoring data, and combined with chaotic characteristic parameters and other relevant monitoring data, comprehensively judge the possibility and severity of potential fault risks. In the control and display management unit, the control instruction generation module sets corresponding threshold intervals based on the key information such as the remaining service life of the combustion chamber predicted by the data comprehensive analysis unit and the fault root cause traced by the fault root cause tracing and data analysis unit. Considering the key role of the combustion chamber in the gas turbine generator set, the maintenance reminder threshold is set at about 25% of the remaining service life. When the remaining service life of the combustion chamber is lower than this maintenance reminder threshold, a maintenance reminder instruction is automatically generated to notify the operation and maintenance personnel to conduct a detailed inspection and necessary maintenance work on the combustion chamber. If after tracing the fault root cause and determining that the possibility of the fault root cause is relatively high, a component replacement suggestion instruction is generated to guide the operation and maintenance personnel to replace the relevant components in time to prevent serious faults of the gas turbine generator set caused by the damage of the combustion chamber components. At the same time, combined with the fault root cause tracing result, if it is determined that a certain operating parameter (such as gas flow rate, air intake volume, etc.) is abnormal and causes the fault, an instruction to adjust the operating parameter can also be generated to guide the operation and maintenance personnel to perform the corresponding parameter adjustment operation to restore the normal operation state of the unit. The information display module intuitively displays the microscopic structure evolution of the combustion chamber through display devices such as liquid crystal displays, including visual information such as the crystal structure deformation image of the combustion chamber wall and the development of microscopic cracks, and at the same time displays the specific values of various monitoring parameters, such as dislocation density, microscopic crack initiation rate, etc.And display important information such as the predicted remaining useful life, the root cause tracing results obtained by the root cause tracing and data analysis unit for faults, and the chaotic data analysis results (if any), so that the operation and maintenance personnel can comprehensively and clearly understand the current state of the combustion chamber, the specific conditions related to faults, and potential risks, thereby providing an intuitive reference basis for the operation and maintenance personnel to take corresponding measures.

[0041] In summary, the present invention plays an important role in aspects such as the monitoring, maintenance, fault troubleshooting, and potential risk excavation of the combustion chamber of a gas turbine generator set, effectively improving the ability to control the state of the combustion chamber, providing a strong guarantee for the stable operation of the gas turbine generator set, being able to more accurately respond to possible problems, and reducing the likelihood of faults and their impact on the operation of the unit.

[0042] For those skilled in the art, it is obvious that the present invention is not limited to the details of the above exemplary embodiments, and can be implemented in other specific forms without departing from the spirit or basic characteristics of the present invention. Therefore, from any point of view, the embodiments should be regarded as exemplary and non-limiting. The scope of the present invention is defined by the appended claims rather than the above description. Therefore, it is intended to embrace all changes falling within the meaning and scope of the equivalent elements of the claims in the present invention. Any reference signs in the claims should not be regarded as limiting the claims involved.

Claims

1. A gas turbine generator set automatic monitoring system, characterized in that: The system includes the following components: micro-state monitoring unit, data collection and transmission unit, data comprehensive analysis unit, fault root cause tracing and data analysis unit and control and display management unit; The microscopic state monitoring unit uses an electron microscope probe with high-resolution imaging capability and an array composed of various types of micro-nano sensors as microscopic monitoring equipment. The sensors include but are not limited to strain sensors, capacitance sensors and resistance sensors. For key components of gas turbine generator sets, key areas prone to microstructural changes are determined based on component structure, operating conditions and experience, and the monitoring equipment is laid out and installed; The data collection and transmission unit is equipped with a multi-channel high-precision data collection card, which can connect to the image and electrical signal data output by the microscopic monitoring unit. For image data, a specific collection frame rate is set according to the monitoring accuracy and processing capacity to ensure the capture of continuous changes in the microstructure of key components. For electrical signal data, the sampling frequency and resolution collection parameters are set according to the output characteristics of different sensors. By adopting optical fiber communication and wireless communication, the advantages of each are utilized to enable the data to be quickly transmitted from the collection card to the comprehensive analysis unit under different working conditions; The data comprehensive analysis unit includes three modules: image information processing, electrical signal data processing and life prediction. The image information processing module pre-processes the received image, and identifies and quantifies the key features related to the microstructure changes of the component based on the algorithm of the convolutional neural network. The electrical signal data processing module restores the electrical signal to the parameter value related to the actual physical and chemical property changes according to the sensor output characteristics and the calibration model. The life prediction module combines the material science related theoretical model with the big data analysis model, takes the microstructure evolution data extracted and restored from the above two modules as input, and predicts the remaining service life of the key components after complex calculations, providing a basis for maintenance decisions; The fault root cause tracing and data analysis unit includes a knowledge graph construction and fault root cause tracing subunit, a chaos data analysis and potential fault risk mining subunit, a chaos data analysis and potential fault risk mining subunit and a cross-domain knowledge transfer subunit, wherein the knowledge graph construction and fault root cause tracing subunit includes a knowledge graph construction module and a fault root cause tracing module, the knowledge graph construction module collects various types of information in the unit field and sorts and classifies them, constructs a knowledge graph according to the inherent logical relationship of knowledge, and clarifies the various associations between entities and relationships, the fault root cause tracing module takes the fault warning data as the starting point, uses the knowledge graph for reverse reasoning, determines the relevant entities and relationships through intelligent algorithms and traces them back, investigates the causes to determine the root causes of the faults, and the chaos data analysis and potential fault risk mining subunit includes a chaos feature parameter calculation module and an operation status judgment and potential fault Risk mining module, chaos characteristic parameter calculation module determines parameters for received data and uses corresponding algorithms to calculate chaos characteristic parameters. Operation status judgment and potential fault risk mining module judges the unit operation status based on these parameters, mines potential fault risks, and comprehensively judges the possibility and severity of risks in combination with other monitoring data. The cross-domain knowledge transfer subunit includes knowledge combing and extraction module, mapping relationship establishment module and knowledge fusion module. The knowledge combing and extraction module collects data from other industrial equipment fields and organizes and extracts knowledge on fault modes, operation parameter changes and maintenance measures. The mapping relationship establishment module compares the knowledge content of the two fields, calculates the correlation and establishes an accurate mapping relationship. The knowledge fusion module integrates the source domain knowledge into the knowledge graph of the gas turbine generator set based on this relationship, thereby improving the accuracy of fault root cause tracing and potential risk mining. The control and display management unit includes a control instruction generation module and an information display module, wherein the control instruction generation module sets a threshold range based on the remaining service life prediction results of the key components of the data comprehensive analysis unit and the fault root cause tracing and fault root source information traced by the data analysis unit, and the information display module intuitively displays the microstructure evolution of the key components, various monitoring parameters, predicted remaining service life, fault root cause tracing results and chaotic data analysis results information through a display device.

2. The automatic monitoring system for a gas turbine generator set according to claim 1, characterized in that: In the data collection and transmission unit, for image data, a specific collection frame rate is set according to the monitoring accuracy and processing capacity, and the image data frame rate optimization formula is adopted. To determine the frame rate, V c is the characteristic speed of microstructural change of key components, L s is the characteristic length scale of microstructural changes in key components, D v Indicates the effective utilization of data transmission bandwidth, T p It is the available storage time period of the data storage device taking into account the capacity of the data storage device and the data storage strategy.

3. The automatic monitoring system for a gas turbine generator set according to claim 1, characterized in that: The data collection and transmission unit adopts both optical fiber communication and wireless communication. In terms of optical fiber communication, optical fiber communication is selected as the main transmission means. In terms of wireless communication, for special occasions such as rotating turbine blades where wiring is inconvenient, ZigBee or Wi-Fi communication is adopted. By setting wireless transmission nodes near the components, the collected data can be wirelessly transmitted to the data comprehensive analysis unit.

4. The automatic monitoring system for a gas turbine generator set according to claim 1, characterized in that: In the data comprehensive analysis unit, the image information processing module preprocesses the received image and identifies and quantifies key features related to the microstructure changes of the component based on the convolutional neural network algorithm. The algorithm formula is: Among them, Q f Represents the quantitative value of a microstructure feature in the image, w i is the weight of the i-th neuron associated with this feature in the CNN model, A i is the activation value of the ith neuron, c is the bias term, and N represents the total number of neurons associated with the microstructural feature.

5. The automatic monitoring system for a gas turbine generator set according to claim 1, characterized in that: In the data comprehensive analysis unit, the life prediction module combines the material science related theoretical model with the big data analysis model, takes the microstructure evolution data extracted and restored from the above two modules as input, and predicts the remaining service life of key components after complex calculations. The calculation formula is: Among them, L r represents the remaining service life of key components, L0 is the initial design service life of key components, α i is the weight coefficient of the i-th characteristic factor related to the microstructure evolution, F i is the actual value of the i-th characteristic factor related to the microstructure evolution at the current moment, β j is the weight coefficient of the jth characteristic factor related to the macro-operation parameters, M j is the actual value of the jth characteristic factor related to the macro-operation parameters at the current moment, γ k is the weight coefficient of the kth characteristic factor related to other influencing factors, C k It is the actual value of the kth characteristic factor related to other influencing factors at the current moment, I represents the total number of characteristic factors related to microstructural evolution, J represents the total number of characteristic factors related to macroscopic operating parameters, and K represents the total number of characteristic factors related to other influencing factors. Through comprehensive analysis and calculation, the remaining service life of key components can be predicted.

6. The automatic monitoring system for a gas turbine generator set according to claim 1, characterized in that: In the fault root cause tracing and data analysis unit, the fault root cause tracing module takes the fault warning data as the starting point, uses the knowledge graph for reverse reasoning, determines the relevant entities and relationships through intelligent algorithms, and traces back in reverse, investigates the causes, and determines the root cause of the fault. The algorithm formula is: Among them, PF root |E1,E2,…,E n Indicates that a series of related events E1, E2, ..., E n In the event of a fault, the root cause is F root The conditional probability, PE i |F root The root cause of the fault is F τoot Under the condition of i Conditional probability of occurrence, PF root Is the root cause of the fault F root The prior probability of occurrence, PE i |F j The root cause of the fault is F j Under the condition of i Conditional probability of occurrence, PF j Other fault sources F j The prior probability of its occurrence, n represents the number of associated events related to the current fault manifestation, and m represents the number of all probabilistic fault root causes.

7. The automatic monitoring system for a gas turbine generator set according to claim 1, characterized in that: In the fault root cause tracing and data analysis unit, the chaos characteristic parameter calculation module determines parameters for the received data and calculates the chaos characteristic parameters using a corresponding algorithm, and the algorithm formula is: Among them, C state The chaos degree index of the unit operation status, λ max is the maximum value among the calculated Lyapunov exponents, λ i is the i-th Lyapunov exponent, It is used to comprehensively consider the changes in the unit's operating status in all directions. c is the currently calculated correlation dimension, D c It reflects the current complexity of the system under the unit operation state, D min is the minimum value of the correlation dimension under normal operating conditions. c Less than D min When the unit is running abnormally, D max is the maximum value of the correlation dimension in the chaotic state. It is also obtained by analyzing a large number of gas turbine generator sets in the state where chaos may occur. c Less than D min When , it means that the unit is in a chaotic state, and N represents the number of calculated Lyapunov exponents.

8. The automatic monitoring system for a gas turbine generator set according to claim 1, characterized in that: In the fault root cause tracing and data analysis unit, the mapping relationship establishment module and the knowledge fusion module compare the knowledge content of the two fields, calculate the correlation and establish an accurate mapping relationship. The correlation calculation formula is: Among them, R ij represents the correlation between the i-th knowledge element in the source domain knowledge and the j-th knowledge element in the gas turbine generator domain knowledge, w k is the weight coefficient of the kth feature dimension, S ik is the feature value of the i-th knowledge element in the source domain knowledge on the k-th feature dimension, S jk is the eigenvalue of the jth knowledge element in the field knowledge of gas turbine generator sets on the kth feature dimension, and K represents the total number of feature dimensions used to measure the knowledge association.

Images