Primary helium fan health state assessment method based on multi-sensor data fusion

Abstract

The invention discloses a primary helium fan health state evaluation method based on multi-sensor data fusion. According to the method, a vibration sensor, a temperature sensor, a current sensor and the like are used for comprehensively monitoring the primary helium fan, a deep learning model is used for carrying out feature extraction and fusion on multi-source heterogeneous data, a fault diagnosis knowledge base based on a knowledge graph is constructed, and real-time and accurate evaluation of the health state of the primary helium fan and intelligent diagnosis of a potential fault mode are achieved. Meanwhile, a reinforcement learning algorithm is introduced in the method, the maintenance strategy of the main helium fan is automatically optimized, the equipment reliability is improved to the maximum extent, and the whole life cycle cost is reduced. The method provided by the invention has good adaptability and expandability, can be popularized and applied to nuclear power and other industrial fields, and provides important technical support for improving the intelligent level and operation safety of key equipment.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent monitoring and fault diagnosis of nuclear power equipment, and in particular to a main helium blower health status assessment method based on multi-sensor data fusion. Background Art

[0002] The main helium blower is an important auxiliary equipment of nuclear power plants, and its reliable operation is directly related to the safety and economy of nuclear power units. The main helium blower usually uses electromagnetic bearings to support the rotor, achieving high-speed operation without lubrication and wear, greatly improving the operating efficiency and service life of the equipment. However, the working environment of electromagnetic bearings is complex, involving the interaction of multiple physical fields such as electromagnetics, mechanics, and thermals. Its operating state is easily affected by external disturbances and internal parameter drift, resulting in a series of faults such as abnormal vibration, shaft displacement changes, and electromagnetic coil overheating, which seriously threaten the safe operation of the main helium blower.

[0003] Traditional main helium blower status monitoring mainly relies on a single type of sensor, such as vibration sensor, temperature sensor, etc. This single monitoring method is difficult to fully reflect the actual operating status of the equipment, especially under complex working conditions, the measurement results of a single sensor are often not accurate and reliable, and easily lead to missed reports or false reports. In addition, traditional fault diagnosis methods mainly rely on empirical thresholds or simple mathematical and statistical models, lack of in-depth exploration of the physical mechanism of the equipment and the law of fault evolution, and are difficult to adapt to the variability and uncertainty of the main helium blower operating conditions. The accuracy and real-time performance of the diagnostic results cannot meet actual needs.

[0004] In recent years, with the rapid development of sensor technology, artificial intelligence and other fields, intelligent diagnostic methods such as multi-sensor data fusion and machine learning have received more and more attention. By deploying various types of sensors on the main helium blower, collecting monitoring data reflecting different physical quantities, using data-driven modeling and learning algorithms, mining the inherent correlation patterns and evolution laws of the data, and building a highly reliable equipment health status assessment and fault diagnosis model, it provides more comprehensive and accurate decision support for the reliable operation of the main helium blower, which has become the current research hotspot and development trend.

[0005] However, there are still many theoretical and technical challenges in how to effectively integrate multi-source heterogeneous sensor data, extract key health status indicators, build an efficient and reliable fault diagnosis knowledge base, and form a health management strategy for the entire life cycle of the equipment. At present, there is still a lack of mature and complete intelligent monitoring and diagnosis systems for main helium blowers, and systematic research and innovation are urgently needed. Summary of the invention

[0006] In view of the above existing problems, the present invention is proposed.

[0007] Therefore, the present invention provides a main helium fan health status assessment method based on multi-sensor data fusion. The technical scheme of the present invention is as follows: a main helium fan health status assessment method based on multi-sensor data fusion, comprising: installing an eddy current displacement sensor for monitoring axial and radial vibrations at the bearing of the main helium fan to obtain multi-source monitoring data to form a multi-source heterogeneous data set; adopting a multimodal data fusion method based on deep learning to extract comprehensive characteristic indicators reflecting the health status of the main helium fan; constructing a diagnostic knowledge base for typical faults of the main helium fan, continuously collecting monitoring data during operation, and using a machine learning algorithm to realize online updating and optimization of the knowledge base; applying the knowledge base to perform fault diagnosis on the real-time monitoring data of the main helium fan, quickly identifying the fault type of the equipment through data-driven fault pattern matching, and forming an intelligent management plan for the entire life cycle of the equipment; continuously optimizing health management decisions through reinforcement learning methods to realize dynamic closed-loop optimization of health status assessment and maintenance management decisions of the main helium fan.

[0008] As a preferred solution of the main helium blower health status assessment method based on multi-sensor data fusion described in the present invention, wherein: the acquisition of multi-source monitoring data includes installing eddy current displacement sensors at the bearings of the main helium blower to monitor the vibration state of the bearings, arranging sensors at the radial and axial measuring points of the bearing housing respectively, and monitoring the rotor vibration in all directions. The sensor measurement accuracy is better than 1 μm, and high-quality data reflecting the bearing vibration state is obtained. The sensor output and the bearing vibration displacement satisfy a linear relationship:

[0009] U o =K s ·x+U o

[0010] Among them, U o Expressed as sensor output voltage, K s is the sensitivity, x is the displacement, U o is the zero voltage;

[0011] In addition to vibration data, it also includes physical quantity monitoring data of current, temperature, and bearing magnetic field strength during the operation of the main helium blower. The data acquisition frequency of each sensor is uniformly set to 10kHz, and synchronous sampling is achieved through the data collector to provide a time-aligned data set for multimodal data fusion analysis.

[0012] As a preferred solution of the main helium fan health status assessment method based on multi-sensor data fusion described in the present invention, wherein: the extraction of comprehensive feature indicators reflecting the health status of the main helium fan includes, for the collected multi-source monitoring data, a multi-modal fusion method based on deep learning is used to extract comprehensive features reflecting the health status of the equipment, and the time series data including vibration, temperature, and current are input into 1D CNN for local feature extraction; the structured data including spectrum and axis trajectory diagram are input into 2D CNN for spatial feature extraction; the feature vectors extracted by CNN are cascaded and input into the LSTM network to model the association between the time series features, and the fully connected network is connected to the LSTM output layer to realize the nonlinear fusion of multi-modal features, so as to generate a comprehensive indicator reflecting the health status of the equipment;

[0013] The core calculation process of 1D CNN is:

[0014]

[0015] in, is the jth convolution kernel output of the ith input of the lth layer, w and b are the convolution kernel parameters and bias terms respectively.

[0016] As a preferred solution of the main helium fan health status assessment method based on multi-sensor data fusion described in the present invention, wherein: the fully connected network connected at the LSTM output layer includes, the core calculation process of LSTM is,

[0017] f t =σ(W f ·[h t-1 ,x t ]+b′ f )

[0018] i t =σ(W i ·[h t-1 ,x t ]+b' i )

[0019] o t =σ(W o ·[h t-1 ,x t ]+b' o )

[0020]

[0021] h t =o t ⊙tanh(C t )

[0022] Among them, f t 、it , o t They are respectively the forget gate, input gate, and output gate, C t is the memory unit state, is the candidate state; h t is the hidden state output, W and b' are the parameters to be learned, σ is the Sigmoid activation function, tanh is the hyperbolic tangent activation function; ⊙ represents element-wise multiplication.

[0023] As a preferred solution of the main helium fan health status assessment method based on multi-sensor data fusion described in the present invention, wherein: the construction of a diagnostic knowledge base for typical faults of the main helium fan includes, for the typical fault modes of the main helium fan, constructing a diagnostic knowledge base containing multi-source monitoring data features, the knowledge base uses an ontology model to represent the semantic association between the fault mode and the data feature, and introduces a probabilistic framework to represent the uncertainty of knowledge, supporting continuous learning and updating of the knowledge base;

[0024] The core concepts of the ontology model include: failure modes, components, physical quantities, and data characteristics;

[0025] The semantic relationships between concepts include:

[0026] Fault hasCausedBy Component

[0027] Fault hasSymptom DataFeature

[0028] Component hasPhysicalQuantity PhysicalQuantity

[0029] DataFeature belongTo PhysicalQuantity

[0030] On the basis of the ontology model, probabilistic representation is introduced to formalize the knowledge base into a probabilistic ontology model. Suppose the concept set in the ontology model is:

[0031]

[0032] The set of semantic relations is:

[0033]

[0034] Fault Concept i In the Knowledge Base The prior probability under is calculated by the probability of the relevant components:

[0035]

[0036] in, Indicated in the knowledge base Next, given a component j Fault i The conditional probability of , obtained through expert knowledge or statistical data;

[0037] For data features DataFeature k In the Knowledge Base The marginal probability under , is calculated by the conditional probability of the relevant physical quantities and the concept of fault:

[0038]

[0039] in, Indicated in the knowledge base Under the given physical quantity PhysicalQuantity l DataFeature k The conditional probability of Indicated in the knowledge base Next, given a fault Fault i and PhysicalQuantity l DataFeature k The conditional probability of is estimated through expert knowledge or historical data.

[0040] As a preferred solution of the main helium blower health status assessment method based on multi-sensor data fusion described in the present invention, wherein: the fault diagnosis includes, based on the above-mentioned probabilistic ontology model, given the feature vector x=[X1,…,X K ], the reasoning process of fault diagnosis is expressed as,

[0041]

[0042] Set each data feature to be independent of each other, and the symptom feature of the fault is only related to the corresponding physical quantity PhysicalQuantity l Related, further get:

[0043]

[0044] Troubleshooting results Corresponding to the fault concept with the largest posterior probability

[0045]

[0046] Based on the above knowledge base, given the feature vector x of online monitoring data, the reasoning process of fault diagnosis is:

[0047]

[0048] By solving the conditional probability parameter formula in the knowledge base, the posterior probability of each fault mode is obtained, and the fault mode with the highest probability is the diagnosis result:

[0049]

[0050] The probability parameters in the knowledge base are initialized through expert experience and continuously iterated and updated according to the diagnostic feedback results to achieve self-adaptation between the knowledge base and the actual working conditions.

[0051] As a preferred solution of the main helium blower health status assessment method based on multi-sensor data fusion described in the present invention, wherein: the continuous optimization of health management decisions by reinforcement learning method includes, during the entire operation cycle of the main helium blower, continuously optimizing health management decisions by reinforcement learning method, including preventive maintenance planning, maintenance method selection, spare parts procurement plan, maximizing the life cycle value of the equipment, including state space, action space, reward function r(s,a);

[0052] The interaction process between the reinforcement learning agent and the environment is described as a Markov decision process:

[0053] V π (s) = Eπ[∑t = 0 T γ t r(s t ,π(s t ))|s0=s]

[0054] Among them, V π (s) is the expected cumulative reward of state s, π(s) is the action selected by the decision agent based on strategy π in state s, γ is the reward decay factor, T is time, and E is the expected value of strategy π;

[0055] The optimization goal of the current process is to find the optimal decision strategy π^ to maximize the expected cumulative reward:

[0056] π^=arg max π V π (s)

[0057] Expert experience is used to guide the agent's exploration process and accelerate the convergence of strategy learning.

[0058] Another object of the present invention is to provide a main helium blower health status assessment system based on multi-sensor data fusion, which can significantly improve the intelligence level of main helium blower fault diagnosis and health management, and provide important technical support for the safe and efficient operation of key nuclear power equipment.

[0059] As a preferred solution of the main helium blower health status assessment system based on multi-sensor data fusion described in the present invention, it includes: a data acquisition unit, a data fusion unit, a health status assessment unit, and a knowledge base module;

[0060] The data acquisition unit is responsible for collecting vibration, temperature, and current multi-source sensor data from the main helium blower to provide original input for subsequent data fusion and health status assessment;

[0061] The data fusion unit receives the multi-source heterogeneous data output by the data acquisition unit, performs preprocessing and fusion, and uses a deep learning method to extract key features that can reflect the health status of the equipment from the fused data, providing a more high-level and abstract feature representation for health status assessment;

[0062] The health status assessment unit includes three modules:

[0063] The health assessment module scores and quantifies the overall health status of the main helium blower based on the features extracted by the data fusion unit;

[0064] A fault diagnosis module, which uses the diagnostic rules in the knowledge base module to identify and diagnose the potential fault modes of the main helium blower;

[0065] The decision optimization module comprehensively considers the results of health assessment and fault diagnosis, and uses intelligent optimization algorithms such as reinforcement learning to automatically generate the optimal maintenance decision plan for the main helium blower, including maintenance time and maintenance method;

[0066] The knowledge base module stores typical failure modes, failure characteristics and corresponding diagnosis rules of the main helium blower, builds a knowledge base by comprehensively utilizing expert experience knowledge and historical operation data, and realizes automatic updating and optimization of knowledge through machine learning. The fault diagnosis module in the health status assessment unit directly calls the diagnosis rules in the knowledge base to realize knowledge-based intelligent diagnosis reasoning.

[0067] A computer device comprises a memory and a processor, wherein the memory stores a computer program, and is characterized in that when the processor executes the computer program, a method for evaluating the health status of a main helium blower based on multi-sensor data fusion is implemented.

[0068] A computer-readable storage medium having a computer program stored thereon, characterized in that when the computer program is executed by a processor, the steps of a main helium blower health status assessment method based on multi-sensor data fusion are implemented.

[0069] Beneficial effects of the present invention: The present invention adopts a multi-sensor data fusion method, making full use of multi-source heterogeneous monitoring data such as vibration, temperature, and current, and comprehensively characterizes the operating status of the main helium blower from the perspective of multiple physical quantities. At the same time, a deep learning model is introduced to extract and fuse features from multi-source data, mining out more high-level, abstract and key health status indicators, greatly improving the accuracy and reliability of health status assessment, and effectively overcoming the limitations of traditional single monitoring methods.

[0070] The present invention constructs a knowledge graph-based main helium blower fault diagnosis knowledge base, combines expert experience knowledge with data-driven fault characteristics, and forms a set of complete and explainable fault diagnosis reasoning rules. By matching the real-time collected monitoring data with the knowledge base online, the potential fault mode and severity of the main helium blower can be quickly and accurately identified, realizing the intelligent and real-time fault diagnosis, and providing a strong guarantee for the safe operation of the equipment.

[0071] The present invention combines the results of health status assessment and fault diagnosis with equipment maintenance management decisions, and introduces a reinforcement learning algorithm to automatically optimize the maintenance strategy of the main helium blower. By considering multiple goals such as equipment health status, fault risk, and maintenance cost, autonomous learning and optimization of maintenance timing, methods, and resource allocation can maximize the availability and reliability of equipment, reduce operating costs throughout the life cycle, and achieve intelligent and economical equipment management.

[0072] The main helium blower health status assessment method proposed in the present invention is universal and adaptable, and has good applicability to main helium blowers of different types and power levels. At the same time, the method adopts a modular and open system architecture design, has good scalability, and can flexibly integrate new monitoring methods, diagnostic models and optimization algorithms to adapt to the continuous development of main helium blower technology. In addition, the multi-sensor data fusion, knowledge graph construction, reinforcement learning optimization and other technologies of the present invention can also be extended and applied to other types of nuclear power key equipment and even a wider range of industrial fields, and has broad application prospects and promotion value. BRIEF DESCRIPTION OF THE DRAWINGS

[0073] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work. Among them:

[0074] Figure 1 A schematic flow chart of a method for evaluating the health status of a main helium blower based on multi-sensor data fusion is provided as an embodiment of the present invention.

[0075] Figure 2 A schematic diagram of eddy current sensor installation for a main helium blower health status assessment method based on multi-sensor data fusion provided by an embodiment of the present invention.

[0076] Figure 3 A multi-modal fusion fault diagnosis flow chart of a main helium blower health status assessment method based on multi-sensor data fusion is provided in accordance with an embodiment of the present invention.

[0077] Figure 4 An embodiment of the present invention provides a main helium blower health status assessment method based on multi-sensor data fusion.

[0078] Figure 5 A module schematic diagram of a main helium blower health status assessment system based on multi-sensor data fusion provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0079] In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in the art without creative work should fall within the scope of protection of the present invention.

[0080] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0081] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor is it a separate or selective embodiment that is mutually exclusive with other embodiments.

[0082] The present invention is described in detail with reference to schematic diagrams. When describing the embodiments of the present invention, for the sake of convenience, the cross-sectional diagrams showing the device structure will not be partially enlarged according to the general scale, and the schematic diagrams are only examples, which should not limit the scope of protection of the present invention. In addition, in actual production, the three-dimensional dimensions of length, width and depth should be included.

[0083] At the same time, in the description of the present invention, it should be noted that the directions or positional relationships indicated by the terms "upper, lower, inner and outer" are based on the directions or positional relationships shown in the drawings, which are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific direction, be constructed and operated in a specific direction, and therefore cannot be understood as limiting the present invention. In addition, the terms "first, second or third" are only used for descriptive purposes and cannot be understood as indicating or implying relative importance.

[0084] In the present invention, unless otherwise clearly specified and limited, the terms "install, connect, connect" should be understood in a broad sense, for example: it can be a fixed connection, a detachable connection or an integral connection; it can also be a mechanical connection, an electrical connection or a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0085] Example 1, reference Figure 1 - Figure 4 , which is the first embodiment of the present invention, provides a main helium blower health status assessment method based on multi-sensor data fusion, comprising:

[0086] S1: Install eddy current displacement sensors for monitoring axial and radial vibrations at the bearings of the main helium blower to obtain multi-source monitoring data and form a multi-source heterogeneous data set.

[0087] Furthermore, the acquisition of multi-source monitoring data includes installing eddy current displacement sensors at the main helium blower bearing to monitor the vibration state of the bearing, arranging sensors at the radial and axial measuring points of the bearing housing to monitor the rotor vibration in all directions, and the sensor measurement accuracy is better than 1μm to obtain high-quality data reflecting the vibration state of the bearing. The sensor output and the bearing vibration displacement satisfy a linear relationship:

[0088] U o =K s ·x+U o

[0089] Among them, U o Expressed as sensor output voltage, K s is the sensitivity, x is the displacement, Uo is the zero voltage;

[0090] In addition to vibration data, it also includes other physical quantity monitoring data such as current, temperature, and bearing magnetic field strength collected during the operation of the main helium blower. The data acquisition frequency of each sensor is uniformly set to 10kHz, and synchronous sampling is achieved through the data collector to provide a time-aligned data set for multimodal data fusion analysis.

[0091] It should be noted that the technical indicators of the eddy current sensor meet the following requirements:

[0092] - Range: ±2mm

[0093] -Sensitivity: 8mV / μm

[0094] - Resolution: 0.1μm

[0095] -Frequency response: 0-10kHz

[0096] S2: A multimodal data fusion method based on deep learning is used to extract comprehensive characteristic indicators reflecting the health status of the main helium blower.

[0097] Furthermore, the extraction of comprehensive characteristic indicators reflecting the health status of the main helium blower includes, for the collected multi-source monitoring data, extracting comprehensive features reflecting the health status of the equipment based on a multimodal fusion method of deep learning, inputting time series data including vibration, temperature, and current into 1D CNN for local feature extraction; inputting structured data including spectrum and axis trajectory diagram into 2D CNN for spatial feature extraction; cascading the feature vectors extracted by CNN, inputting them into the LSTM network to model the association between time series features, connecting the fully connected network at the LSTM output layer, realizing nonlinear fusion of multimodal features, and generating comprehensive indicators reflecting the health status of the equipment;

[0098] The core calculation process of the 1D CNN is:

[0099]

[0100] in, is the jth convolution kernel output of the ith input of the lth layer, w and b are the convolution kernel parameters and bias terms respectively.

[0101] It should be noted that the fully connected network at the LSTM output layer includes the core calculation process of LSTM:

[0102] f t =σ(W f ·[h t-1 ,x t ]+b′ f )

[0103] i t =σ(W i ·[h t-1 ,x t ]+b' i )

[0104] o t =σ(W o ·[h t-1 ,x t ]+b' o )

[0105]

[0106] h t =o t ⊙tanh(C t )

[0107] Among them, f t 、i t , o t They are respectively the forget gate, input gate, and output gate, C t is the memory unit state, is the candidate state; h t is the hidden state output, W and b' are the parameters to be learned, σ is the Sigmoid activation function, tanh is the hyperbolic tangent activation function; ⊙ represents element-wise multiplication.

[0108] S3: Build a diagnostic knowledge base for typical faults of the main helium blower, continuously collect monitoring data during operation, and use machine learning algorithms to achieve online update and optimization of the knowledge base.

[0109] Furthermore, the construction of the diagnosis knowledge base of the typical faults of the main helium blower includes, for the typical fault modes of the main helium blower, constructing a diagnosis knowledge base containing multi-source monitoring data features, the knowledge base adopts an ontology model to represent the semantic association between the fault mode and the data feature, and introduces a probability framework to represent the uncertainty of knowledge, so as to support the continuous learning and updating of the knowledge base;

[0110] The core concepts of the ontology model include:

[0111] -Fault mode: such as imbalance, friction, looseness, etc.

[0112] -Component: such as rotor, bearing, impeller, etc.;

[0113] -Physical Quantity: such as vibration, temperature, current, etc.;

[0114] -Data Features: such as spectrum peak, root mean square, etc.

[0115] The semantic relationships between concepts include:

[0116] FaulthasCausedBy Component

[0117] FaulthasSymptom DataFeature

[0118] ComponenthasPhysicalQuantity PhysicalQuantity

[0119] DataFeature belongTo PhysicalQuantity

[0120] On the basis of the ontology model, probabilistic representation is introduced to formalize the knowledge base into a probabilistic ontology model. Suppose the concept set in the ontology model is:

[0121]

[0122] The set of semantic relations is:

[0123]

[0124] Fault Concept i In the Knowledge Base The prior probability under is calculated by the probability of the relevant components:

[0125]

[0126] in, Indicated in the knowledge base Next, given a component j Fault i The conditional probability of is obtained through expert knowledge or statistical data.

[0127] For data features DataFeature k In the Knowledge Base The marginal probability under , is calculated by the conditional probability of the relevant physical quantities and the concept of fault:

[0128]

[0129] in, Indicated in the knowledge base Under the given physical quantity PhysicalQuantity l DataFeature k The conditional probability of Indicated in the knowledge base Next, given a fault Fault i and PhysicalQuantity l DataFeature k The conditional probability of is estimated through expert knowledge or historical data.

[0130] S4: Apply the knowledge base to perform fault diagnosis on the real-time monitoring data of the main helium blower, quickly identify the fault type of the equipment through data-driven fault pattern matching, and form an intelligent management plan for the entire life cycle of the equipment.

[0131] Furthermore, the fault diagnosis includes, based on the above-mentioned probabilistic ontology model, given the feature vector x=[X1,…,X K ], the reasoning process of fault diagnosis is expressed as,

[0132]

[0133] Set each data feature to be independent of each other, and the symptom feature of the fault is only related to the corresponding physical quantity PhysicalQuantity l Related, further get:

[0134]

[0135] Troubleshooting results Corresponding to the fault concept with the largest posterior probability

[0136]

[0137] Based on the above knowledge base, given the feature vector x of online monitoring data, the reasoning process of fault diagnosis is:

[0138]

[0139] By solving the conditional probability parameter formula in the knowledge base, the posterior probability of each fault mode is obtained, and the fault mode with the highest probability is the diagnosis result:

[0140]

[0141] The probability parameters in the knowledge base are initialized through expert experience and continuously iterated and updated according to the diagnostic feedback results to achieve self-adaptation between the knowledge base and the actual working conditions.

[0142] S5: Continuously optimize health management decisions through reinforcement learning methods to achieve dynamic closed-loop optimization of health status assessment and maintenance management decisions of the main helium blower.

[0143] Furthermore, the continuous optimization of health management decisions through reinforcement learning methods includes, during the entire operation cycle of the main helium blower, continuous optimization of health management decisions through reinforcement learning methods, including preventive maintenance planning, maintenance method selection, spare parts procurement plan, maximizing the life cycle value of the equipment, mainly including:

[0144] -State space: Integrates information such as equipment health status, failure risk, historical maintenance records, spare parts inventory, etc. to form a multidimensional state vector s;

[0145] -Action space: including discrete or continuous actions such as maintenance start time, maintenance method, spare parts procurement, etc.a;

[0146] - Reward function: Considering factors such as equipment reliability, maintenance cost, and spare parts optimization, a reward function r(s,a) is designed to quantify the value of the equipment throughout its life cycle;

[0147] The interaction process between the reinforcement learning agent and the environment is described as a Markov decision process:

[0148] V π (s) = Eπ[∑t = 0 T γ t r(s t ,π(s t ))|s0=s]

[0149] Among them, V π (s) is the expected cumulative reward of state s, π(s) is the action selected by the decision agent based on strategy π in state s, γ is the reward decay factor, T is time, and E is the expected value of strategy π;

[0150] The optimization goal of the current process is to find the optimal decision strategy π^ to maximize the expected cumulative reward:

[0151] π^=arg max π V π (s)

[0152] Expert experience is used to guide the agent's exploration process and accelerate the convergence of strategy learning.

[0153] Example 2, the first example of the present invention, provides a main helium blower health status assessment method based on multi-sensor data fusion. In order to verify the beneficial effects of the present invention, scientific demonstration is carried out through experiments.

[0154] First, vibration sensors, temperature sensors and current sensors are installed on the key parts of the experimental bench fan (such as bearings, impellers, motors, etc.) to achieve comprehensive monitoring of the equipment's operating status.

[0155] Secondly, a large amount of historical operating data and fault cases of the main helium blower were collected through experiments, and these data were analyzed and annotated. On this basis, we built a fault diagnosis knowledge base containing five typical fault modes (such as imbalance, friction, mechanical looseness, etc.), and used the deep learning fusion method of the present invention to achieve automatic optimization and update of the diagnostic rules in the knowledge base.

[0156] During the experiment, we used the fault diagnosis knowledge base we built to conduct online analysis of multi-sensor monitoring data collected in real time. When the equipment is in an abnormal state, this method can promptly detect abnormal characteristics and provide potential fault causes and severity assessments, providing a reliable basis for maintenance personnel to make maintenance decisions.

[0157] Through experimental verification, the method of the present invention successfully detected two bearing failures and one impeller failure of the main helium blower, effectively improving the operating reliability of the main helium blower. Compared with the traditional single sensor monitoring and threshold alarm method, the fault detection rate of the method of the present invention increased by 20% and the missed alarm rate decreased by 15%, reflecting significant technical advantages and engineering application value.

[0158] In summary, the main helium blower health status assessment method based on multi-sensor data fusion provided by the present invention can effectively improve the accuracy and timeliness of fault diagnosis and provide important technical support for the reliable operation of the main helium blower. This method makes full use of artificial intelligence technologies such as deep learning, knowledge graphs, and reinforcement learning, and is an important means to achieve intelligent operation and maintenance of the main helium blower, with broad engineering application prospects.

[0159] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

[0160] Embodiment 3, the third embodiment of the present invention, is different from the first two embodiments in that:

[0161] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc., which can store program codes.

[0162] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute instructions), or in conjunction with such instruction execution systems, devices or apparatuses. For the purposes of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in conjunction with such instruction execution systems, devices or apparatuses.

[0163] More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic device), a portable computer disk case (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be a paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering or, if necessary, processing in another suitable manner, and then stored in a computer memory.

[0164] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, a plurality of steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0165] Example 4, reference Figure 5 , which is an embodiment of the present invention, provides a main helium blower health status assessment system based on multi-sensor data fusion, characterized by comprising a data acquisition unit, a data fusion unit, a health status assessment unit, and a knowledge base module.

[0166] The data acquisition unit is responsible for collecting vibration, temperature, and current multi-source sensor data from the main helium blower, providing raw input for subsequent data fusion and health status assessment.

[0167] The data fusion unit receives the multi-source heterogeneous data output by the data acquisition unit, performs preprocessing and fusion, and uses deep learning methods to extract key features that can reflect the health status of the equipment from the fused data, providing a more high-level and abstract feature representation for health status assessment.

[0168] The health status assessment unit consists of three modules:

[0169] The health assessment module scores and quantifies the overall health status of the main helium blower based on the features extracted by the data fusion unit;

[0170] A fault diagnosis module, which uses the diagnostic rules in the knowledge base module to identify and diagnose the potential failure modes of the main helium blower;

[0171] The decision optimization module comprehensively considers the results of health assessment and fault diagnosis, and adopts intelligent optimization algorithms such as reinforcement learning to automatically generate the optimal maintenance decision plan for the main helium blower, including maintenance time and maintenance method.

[0172] The knowledge base module stores the typical failure modes, failure characteristics and corresponding diagnostic rules of the main helium blower. It builds a knowledge base by comprehensively utilizing expert experience knowledge and historical operation data, and realizes automatic updating and optimization of knowledge through machine learning. The fault diagnosis module in the health status assessment unit directly calls the diagnostic rules in the knowledge base to realize knowledge-based intelligent diagnostic reasoning.

[0173] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A method for evaluating the health status of main helium blowers based on multi-sensor data fusion. Features: include, Install eddy current displacement sensors for monitoring axial and radial vibrations at the main helium blower bearings to obtain multi-source monitoring data and form a multi-source heterogeneous data set; A multimodal data fusion method based on deep learning is used to extract comprehensive characteristic indicators reflecting the health status of the main helium blower; Build a diagnostic knowledge base for typical faults of the main helium blower, continuously collect monitoring data during operation, and use machine learning algorithms to achieve online update and optimization of the knowledge base; Apply the knowledge base to perform fault diagnosis on the real-time monitoring data of the main helium blower, quickly identify the fault type of the equipment through data-driven fault pattern matching, and form an intelligent management solution for the entire life cycle of the equipment; Through the reinforcement learning method, health management decisions are continuously optimized to achieve dynamic closed-loop optimization of the health status assessment and maintenance management decisions of the main helium blower.

2. A method for evaluating the health status of a main helium blower based on multi-sensor data fusion as claimed in claim 1, Features: The acquisition of multi-source monitoring data includes installing eddy current displacement sensors at the main helium blower bearings to monitor the vibration state of the bearings, arranging sensors at radial and axial measuring points of the bearing housing to monitor the vibration of the rotor in all directions, and obtaining high-quality data reflecting the vibration state of the bearings with a sensor measurement accuracy better than 1 μm. The sensor output and the bearing vibration displacement satisfy a linear relationship: U o =K s ·x+U o Among them, U o Expressed as sensor output voltage, K s is the sensitivity, x is the displacement, U o is the zero voltage; In addition to vibration data, it also includes physical quantity monitoring data of current, temperature, and bearing magnetic field strength during the operation of the main helium blower. The data acquisition frequency of each sensor is uniformly set to 10kHz, and synchronous sampling is achieved through the data collector to provide a time-aligned data set for multimodal data fusion analysis.

3. A method for evaluating the health status of a main helium blower based on multi-sensor data fusion as claimed in claim 2, Features: The method of extracting comprehensive characteristic indicators reflecting the health status of the main helium blower includes extracting comprehensive characteristics reflecting the health status of the equipment based on a multi-modal fusion method of deep learning for the collected multi-source monitoring data, inputting time series data including vibration, temperature, and current into 1D CNN for local feature extraction; inputting structured data including spectrum and axis trajectory diagram into 2D CNN for spatial feature extraction; cascading the feature vectors extracted by CNN, inputting them into the LSTM network to model the association between the time series features, connecting the fully connected network at the LSTM output layer, realizing nonlinear fusion of multi-modal features, and generating comprehensive indicators reflecting the health status of the equipment; The core calculation process of 1D CNN is: in, is the jth convolution kernel output of the ith input of the first layer, w and b are the convolution kernel parameters and bias terms respectively.

4. A method for evaluating the health status of a main helium blower based on multi-sensor data fusion as claimed in claim 3, Features: The fully connected network at the LSTM output layer includes the core calculation process of LSTM: f t =σ(W f ·[h t-1 ,x t ]+b′ f ) i t =σ(W i ·[h t-1 ,x t ]+b i ) the t =σ(W o ·[h t-1 ,x t +b′ o ) h t =o t ⊙tanh(C t ) Among them, f t 、i t , o t They are respectively the forget gate, input gate, and output gate, C t is the memory unit state, is the candidate state; h t is the hidden state output, W and b′ are the parameters to be learned, σ is the Sigmoid activation function, tanh is the hyperbolic tangent activation function; ⊙ represents element-wise multiplication.

5. A method for evaluating the health status of a main helium blower based on multi-sensor data fusion as claimed in claim 4, Features: The construction of the diagnosis knowledge base of the typical faults of the main helium blower includes: for the typical fault modes of the main helium blower, a diagnosis knowledge base containing multi-source monitoring data features is constructed, the knowledge base uses an ontology model to represent the semantic association between the fault mode and the data feature, and introduces a probability framework to represent the uncertainty of knowledge, so as to support the continuous learning and updating of the knowledge base; The core concepts of the ontology model include: failure modes, components, physical quantities, and data characteristics; The semantic relationships between concepts include: Fault hasCausedBy Component Fault hasSymptom DataFeature Component hasPhysicalQuantity PhysicalQuantity DataFeature belongTo PhysicalQuantity On the basis of the ontology model, probabilistic representation is introduced to formalize the knowledge base into a probabilistic ontology model. Suppose the concept set in the ontology model is: The set of semantic relations is: Fault concept i In the Knowledge Base The prior probability under is calculated by the probability of the relevant components: in, Indicated in the knowledge base Next, given a component j Fault i The conditional probability of , obtained through expert knowledge or statistical data; For data features DataFeature k In the Knowledge Base The marginal probability under , is calculated by the conditional probability of the relevant physical quantities and the fault concept: in, Indicated in the knowledge base Under the given physical quantity PhysicalQuantity l DataFeature k The conditional probability of Indicated in the knowledge base Next, given a fault Fault i and PhysicalQuantity l DataFeature k The conditional probability of is estimated through expert knowledge or historical data.

6. A method for evaluating the health status of a main helium blower based on multi-sensor data fusion as claimed in claim 5, Features: The fault diagnosis comprises: based on the above-mentioned probabilistic ontology model, given the feature vector x=[X 1 ,…,X K ], the reasoning process of fault diagnosis is expressed as, Set each data feature to be independent of each other, and the symptom feature of the fault is only related to the corresponding physical quantity PhysicalQuantity l Related, further get: Troubleshooting results Corresponding to the fault concept with the largest posterior probability Based on the above knowledge base, given the feature vector x of online monitoring data, the reasoning process of fault diagnosis is: By solving the conditional probability parameter formula in the knowledge base, the posterior probability of each fault mode is obtained, and the fault mode with the highest probability is the diagnosis result: The probability parameters in the knowledge base are initialized through expert experience and continuously iterated and updated according to the diagnostic feedback results to achieve self-adaptation between the knowledge base and the actual working conditions.

7. A method for evaluating the health status of a main helium blower based on multi-sensor data fusion as claimed in claim 6, Features: The continuous optimization of health management decisions by reinforcement learning methods includes, during the entire operation cycle of the main helium blower, continuous optimization of health management decisions by reinforcement learning methods, including preventive maintenance planning, maintenance method selection, spare parts procurement plan, maximizing the life cycle value of the equipment, including state space, action space, and reward function r(s,a); The interaction process between the reinforcement learning agent and the environment is described as a Markov decision process: IN π (s)=Eπ[∑t=O T γ t r(s t ,π(s t ))|s 0 [with] Among them, V π (s) is the expected cumulative reward of state s, π(s) is the action selected by the decision agent based on strategy π in state s, γ is the reward decay factor, T is time, and E is the expected value of strategy π; The optimization goal of the current process is to find the optimal decision strategy π^ to maximize the expected cumulative reward: π^=arg max π V π (s) Expert experience is used to guide the agent's exploration process and accelerate the convergence of strategy learning.

8. A system using the main helium blower health status assessment method based on multi-sensor data fusion as claimed in any one of claims 1 to 7, Features: It includes data acquisition unit, data fusion unit, health status assessment unit and knowledge base module; The data acquisition unit is responsible for collecting vibration, temperature, and current multi-source sensor data from the main helium blower to provide original input for subsequent data fusion and health status assessment; The data fusion unit receives the multi-source heterogeneous data output by the data acquisition unit, performs preprocessing and fusion, and uses a deep learning method to extract key features that can reflect the health status of the equipment from the fused data, providing a more high-level and abstract feature representation for health status assessment; The health status assessment unit includes three modules: The health assessment module scores and quantifies the overall health status of the main helium blower based on the features extracted by the data fusion unit; A fault diagnosis module, which uses the diagnostic rules in the knowledge base module to identify and diagnose the potential failure modes of the main helium blower; The decision optimization module comprehensively considers the results of health assessment and fault diagnosis, and uses intelligent optimization algorithms such as reinforcement learning to automatically generate the optimal maintenance decision plan for the main helium blower, including maintenance time and maintenance method; The knowledge base module stores typical failure modes, failure characteristics and corresponding diagnosis rules of the main helium blower, builds a knowledge base by comprehensively utilizing expert experience knowledge and historical operation data, and realizes automatic updating and optimization of knowledge through machine learning. The fault diagnosis module in the health status assessment unit directly calls the diagnosis rules in the knowledge base to realize knowledge-based intelligent diagnosis reasoning.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program. It is characterized in that When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, It is characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images