Method and system for evaluating device health status under fault-free samples

Abstract

The present application relates to the technical field of equipment health state evaluation, and an embodiment provides an equipment health state evaluation method and system under a failure-free sample, wherein the method comprises the following steps: constructing a self-sample training set according to historical data samples of the health state of equipment state data; generating a boundary detector set according to a nearest neighbor fixed boundary negative selection algorithm based on the self-sample training set, wherein the boundary detector set comprises a plurality of boundary detectors; taking equipment state data to be evaluated as a test sample, and determining whether the equipment state data is healthy according to the relationship between the test sample and the boundary detector set. The embodiment provided by the present application can realize the health state evaluation of equipment under a failure-free sample, enrich the application scenarios, and improve the detection efficiency.

Description

Technical Field

[0001] This invention relates to the field of equipment health status assessment technology, specifically to a method for assessing equipment health status under a fault-free sample, a system for assessing equipment health status under a fault-free sample, an electronic device, and a corresponding storage medium. Background Technology

[0002] Mechanical equipment is a crucial component of my country's industrial production. Conducting health status assessments of mechanical equipment is of paramount importance for reducing unplanned downtime and ensuring production safety. Health status assessment of mechanical equipment is developed based on fault diagnosis technology and can be categorized into three methods according to their principles: model-driven, knowledge-driven, and data-driven methods. Model-driven methods achieve status assessment by establishing mathematical models, yielding relatively accurate results, but are not suitable for complex equipment. Knowledge-driven methods do not require the construction of precise mathematical models but demand extensive expert experience. Data-driven methods primarily rely on equipment operating data, requiring less expert knowledge and understanding of mechanisms, but necessitate a certain amount of historical data.

[0003] Thanks to the development of condition monitoring systems and intelligent algorithms, data-driven health status assessment methods have made continuous progress and gained increasing applications. This is mainly due to two reasons: First, with the continuous improvement of modern industrial levels, the normal operation of complex mechanical equipment generally requires the support of multiple auxiliary subsystems, such as lubrication systems and sealing systems, which greatly increases the difficulty of applying both knowledge-based and model-based condition assessment methods. Second, to ensure the safe operation of equipment, complex mechanical equipment is generally equipped with condition monitoring systems, which basically cover key locations of the equipment body and auxiliary subsystems, enabling real-time monitoring of key parameters and providing a solid data foundation for assessing the health status of the equipment.

[0004] Based on whether or not fault data is used, data-driven methods can be further divided into two categories: fault data-based methods and normal data-based methods. Fault data-based methods require training samples in different states (including normal states and different abnormal states), while normal data-based methods only require training samples in normal states. As equipment reliability improves, abnormal equipment samples become scarce, making it difficult to meet the sample requirements of fault data-based early warning methods, thus limiting the practical application of this type of method.

[0005] The negative selection algorithm, based on the biological immune mechanism, can assess the health status of equipment even with fault-free samples. Similar to the immune system's self-non-self recognition mechanism, the basic idea of ​​the negative selection algorithm is to generate a corresponding set of detectors based on the normal dataset of the equipment, and then use these detectors to match the data to be judged, thereby identifying and assessing the status of the data samples. Compared to methods based on fault data, the negative selection algorithm has the advantage of requiring only one type of sample to generate the rule base and can obtain some rules that do not require prior knowledge. Therefore, this method has been widely applied in fields such as network intrusion detection, text classification, fault detection, and disease diagnosis.

[0006] The detector is crucial to the negative selection algorithm, and its generation mechanism is the core technology of the algorithm. Traditional negative selection algorithms generally generate detectors randomly. To improve the coverage of detectors in non-self spaces, a large number of detectors need to be generated, resulting in excessively long computation time and low detection efficiency. In addition, the sample matching rules of traditional negative selection algorithms are also insufficient, especially in the absence of clear sample determination rules in detection blind zones. Summary of the Invention

[0007] The purpose of this invention is to provide a method and system for assessing the health status of equipment under fault-free samples. In view of the shortcomings of traditional negative selection algorithms in detector generation mechanism and sample matching rules, this invention proposes a Nearest Boundary-Fixed Negative Selection Algorithm (NFB-NSA) for equipment health status assessment. It aims to solve the problem of sample identification in the detection blind zone, reduce the number of detectors generated, and thus achieve higher detection efficiency.

[0008] To achieve the above objectives, the first aspect of the present invention provides a method for assessing the health status of equipment under fault-free samples. The method includes: constructing a self-sample training set based on historical data samples of equipment status data as health status; generating a boundary detector set from the self-sample training set using a nearest neighbor fixed boundary negation selection algorithm, wherein the boundary detector set includes a plurality of boundary detectors; using the equipment status data to be assessed as a test sample, and determining whether the equipment status data is healthy based on the relationship between the test sample and the boundary detector set.

[0009] Preferably, constructing a self-sample training set based on historical data samples of health status from device status data includes: determining the data representation method of the historical data samples; determining the representation method of the boundary detector based on the data representation method of the historical data samples; and determining the matching rules between the historical data samples and the boundary detector.

[0010] Preferably, generating a boundary detector set from the self-sample training set using the nearest neighbor fixed boundary negation selection algorithm includes: dividing the state space into m parts based on the features of the self-samples in the self-sample training set. n Each hypercube is assigned a center point; the properties of each hypercube are determined based on its relationship with the self-sample; and based on the properties of each hypercube, the m... n The outer boundary hypercube is determined from the hypercubes; the boundary detector set is determined based on the outer boundary hypercubes.

[0011] Preferably, the properties of each hypercube are determined based on its relationship with the self-sample, including: If a hypercube is covered by the self-sample, or if the self-sample is inside the hypercube, then the hypercube is determined to be a non-empty hypercube; otherwise, the hypercube is determined to be an empty hypercube.

[0012] Preferably, based on the properties of each hypercube, from the m n Determining the outer boundary hypercube among a hypercube includes: if a hypercube is an empty hypercube and at least one of its neighboring hypercubes is a non-empty hypercube, then the hypercube is an outer boundary hypercube.

[0013] Preferably, determining the boundary detector set based on the outer boundary hypercube includes: taking the center of the outer boundary hypercube as the center of the boundary detector, determining the radius of the boundary detector based on the minimum distance to the self-sample, and obtaining the boundary detector corresponding to the outer boundary hypercube; a number of boundary detectors constitute the boundary detector set.

[0014] Preferably, determining whether the device status data is healthy based on the relationship between the sample to be tested and the boundary detector includes: determining that the sample to be tested is not in the detection blind zone of the boundary detector; when the sample to be tested is within the boundary detector, the sample to be tested is determined to be a non-self sample, and the device status data is unhealthy. When the sample to be tested is not within the boundary detector, if the sample to be tested and the self sample in the self sample training set are located on the same side of the boundary detector, then the sample to be tested is determined to be a self sample, and the device status data is healthy. When the sample to be tested is not within the boundary detector, if the sample to be tested and the self sample in the self sample training set are located on opposite sides of the boundary detector, then the sample to be tested is determined to be a non-self sample, and the device status data is unhealthy.

[0015] Preferably, if the sample to be tested is in the detection blind zone of the boundary detector, the sample to be tested is determined to be a self-sample or a non-self-sample by the nearest sample to the sample to be tested, and then the device status data corresponding to the sample to be tested is determined to be healthy or unhealthy.

[0016] In a second aspect of the invention, a system for assessing the health status of equipment under fault-free samples is also provided. The system includes: a training set generation module, configured to construct a self-sample training set based on historical data samples of health status from equipment status data; a detector generation module, configured to generate a boundary detector set from the self-sample training set using a nearest neighbor fixed boundary negation selection algorithm, wherein the boundary detector set includes a plurality of boundary detectors; and a status assessment module, configured to use the equipment status data to be assessed as a test sample, and determine whether the equipment status data is healthy based on the relationship between the test sample and the boundary detector set.

[0017] Preferably, a self-sample training set is constructed based on historical data samples of device status for health status, including: The process involves determining the data representation method of the historical data samples; determining the representation method of the boundary detector based on the data representation method of the historical data samples; and determining the matching rules between the historical data samples and the boundary detector.

[0018] Preferably, generating a boundary detector set from the self-sample training set using the nearest neighbor fixed boundary negation selection algorithm includes: dividing the state space into m parts based on the features of the self-samples in the self-sample training set. n Each hypercube is assigned a center point; the properties of each hypercube are determined based on its relationship with the self-sample; and based on the properties of each hypercube, the m... n The outer boundary hypercube is determined from the hypercubes; the boundary detector set is determined based on the outer boundary hypercubes.

[0019] Preferably, the properties of each hypercube are determined based on the relationship between each hypercube and the self-sample, including: when a hypercube is covered by the self-sample, or when the self-sample is inside the hypercube, the hypercube is determined to be a non-empty hypercube; otherwise, the hypercube is determined to be an empty hypercube.

[0020] Preferably, based on the properties of each hypercube, from the m nDetermining the outer boundary hypercube among a hypercube includes: if a hypercube is an empty hypercube and at least one of its neighboring hypercubes is a non-empty hypercube, then the hypercube is an outer boundary hypercube.

[0021] Preferably, determining the boundary detector set based on the outer boundary hypercube includes: taking the center of the outer boundary hypercube as the center of the boundary detector, determining the radius of the boundary detector based on the minimum distance to the self-sample, and obtaining the boundary detector corresponding to the outer boundary hypercube; a number of boundary detectors constitute the boundary detector set.

[0022] Preferably, determining whether the device status data is healthy based on the relationship between the sample to be tested and the boundary detector includes: determining that the sample to be tested is not in the detection blind zone of the boundary detector; When the sample to be tested is within the boundary detector, the sample to be tested is determined to be a non-self sample, and the device status data is unhealthy. When the sample to be tested is not within the boundary detector, if the sample to be tested and the self sample in the self sample training set are located on the same side of the boundary detector, then the sample to be tested is determined to be a self sample, and the device status data is healthy. When the sample to be tested is not within the boundary detector, if the sample to be tested and the self sample in the self sample training set are located on opposite sides of the boundary detector, then the sample to be tested is determined to be a non-self sample, and the device status data is unhealthy.

[0023] Preferably, if the sample to be tested is in the detection blind zone of the boundary detector, the sample to be tested is determined to be a self-sample or a non-self-sample by the nearest sample to the sample to be tested, and then the device status data corresponding to the sample to be tested is determined to be healthy or unhealthy.

[0024] In a third aspect of the invention, an electronic device is also provided, comprising: at least one processor; and a memory connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the at least one processor implements the aforementioned device health status assessment method under fault-free sample conditions by executing the instructions stored in the memory.

[0025] A fourth aspect of the present invention provides a machine-readable storage medium storing instructions that, when executed by a processor, configure the processor to perform the aforementioned device health status assessment method under a fault-free sample.

[0026] The fifth aspect of the present invention provides a computer program product, including a computer program that, when executed by a processor, implements the above-described method for assessing the health status of equipment under fault-free conditions.

[0027] This invention addresses the problem of scarce equipment fault samples. By introducing a boundary detector, this patent optimizes the detector generation mechanism and matching rules of traditional negative selection algorithms, proposing a nearest neighbor fixed boundary negative selection algorithm (NFB-NSA) for equipment health status assessment. The introduction of a boundary detector changes the traditional detector generation method, avoiding random detector generation, significantly reducing computational load, and helping to maintain result stability. Simultaneously, matching rules for samples in detection blind zones are established, improving detection efficiency. Ultimately, it achieves health status assessment of complex equipment under fault-free sample conditions, thus solving the deficiency of traditional data-driven methods in the current situation of scarce or even missing abnormal data.

[0028] Other features and advantages of the embodiments of the present invention will be described in detail in the following detailed description section. Attached Figure Description

[0029] The accompanying drawings are provided to further illustrate embodiments of the present invention and form part of the specification. They are used together with the following detailed description to explain the embodiments of the present invention, but do not constitute a limitation thereof. In the drawings: Figure 1 The schematic diagram illustrates the steps of a method for assessing equipment health status under a fault-free sample according to an embodiment of the present invention. Figure 2 A schematic flowchart of the nearest neighbor fixed boundary negation selection algorithm according to an embodiment of the present invention is shown. Figure 3 This schematic diagram illustrates the structure of a detector according to the prior art; Figure 4 A schematic diagram of the structure of a boundary detector according to an embodiment of the present invention is shown; Figure 5 The diagram illustrates the generation of a boundary detector in a two-dimensional space according to an embodiment of the present invention. Figure 6 This schematic diagram illustrates the structure of an equipment health status assessment system under a fault-free sample according to an embodiment of the present invention. Figure 7 A schematic diagram of the structure of a centrifugal compressor unit according to an embodiment of the present invention is shown; Figure 8 The diagram illustrates a comparison of early warning results under different training sample sizes according to an embodiment of the present invention. Detailed Implementation

[0030] The specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are for illustration and explanation only and are not intended to limit the scope of the present invention.

[0031] Figure 1 The schematic diagram illustrates the steps of a method for assessing equipment health status under a fault-free sample according to an embodiment of the present invention, such as... Figure 1 As shown. A method for assessing the health status of equipment under a fault-free sample, the method comprising: S01. Construct a self-sample training set based on historical data samples of healthy equipment status data; in the absence or inability to obtain equipment status data of faulty status, use the equipment status data of healthy status as historical data samples to construct the self-sample training set. Each historical data sample is mapped to a self-sample through a preset data representation method. These self-samples are used to obtain detectors, which are used to detect the equipment status data to be detected.

[0032] S02. The self-sample training set is used to generate a boundary detector set using the nearest neighbor fixed boundary negation selection algorithm. This boundary detector set includes several boundary detectors. As the core of the negation selection algorithm, the detector generation mechanism determines the algorithm's accuracy and efficiency. Generating a small number of high-quality detectors in a short time has become crucial for the success of the negation selection algorithm. Similar to the formulation of matching rules, the detector generation mechanism can be divided into two types based on the data and detector representation: string-based generation mechanisms and real-valued vector-based generation mechanisms. String-based generation mechanisms include exhaustive search, linear (greedy) methods, template methods, evolutionary methods, negation databases, and r-variable methods. Real-valued vector-based generation mechanisms include real-valued negation selection algorithms, positive selection algorithms, V-detector algorithms, stepwise methods, adaptive methods, evolutionary methods, and cutting methods.

[0033] S03. Using the device status data to be evaluated as a test sample, determine whether the device status data is normal based on the relationship between the test sample and the boundary detector set.

[0034] In real-valued vector detector generation mechanisms, the V-detector algorithm is a widely used method. This method generates a corresponding set of detectors based on a set of self-samples, and then tries to cover all non-self regions as much as possible. The detector is defined by two parameters: a center point and a radius. The center point is an n-dimensional vector corresponding to a point in the non-self space. The detector is generated using a random method: a center point x is randomly determined in the non-self space, and the distance d between the center point x and the nearest self-sample is determined. A detector with center point x and radius (d-rs) is then generated, where rs is the radius of the nearest self-sample. During the testing phase, if the distance between the sample to be tested and the center point is less than the detector radius, it does not match that detector. If the sample to be tested matches any detector, it is determined to be a non-self sample; otherwise, it is determined to be a self sample. By judging non-self samples and self samples, the health of the sample to be tested is determined.

[0035] Through the above implementation methods, this invention proposes a health status assessment method for complex equipment based on the nearest neighbor fixed boundary negation selection algorithm (NFB-NSA). It is mainly applied to the status assessment of mechanical equipment, electrical equipment and chemical equipment. It can realize the status assessment of equipment without fault samples and is suitable for situations where fault samples are scarce or even missing.

[0036] In some embodiments provided by this invention, a self-sample training set is constructed based on historical data samples where the device health status is normal. This includes: determining the data representation method of the historical data samples; determining the representation method of the boundary detector based on the data representation method of the historical data samples; and determining the matching rules between the historical data samples and the boundary detector. First, the representation of the data and the detector is determined. Data representation is the representation of existing samples and is the first step in the negative selection algorithm, significantly impacting the formulation of subsequent matching rules and the generation of the detector. Data representation needs to be determined based on the analysis object, mainly involving four data types: numerical data, Boolean data, categorical data, and text data. Representation methods can be broadly categorized into three methods: string, real-valued vector, and matrix. The detector representation method is determined according to the data representation method and is also categorized into three methods: string representation, real-valued vector representation, and matrix representation. In the real-valued vector representation method, based on the type of detector model, it can be further divided into hypercube model, hypersphere model, hyperellipsoid model, multi-shape model, etc. The specific representation method of the detector needs to be determined based on the data characteristics. Next, the matching rules are specified. Matching rules, also known as affinity calculation, are used to characterize the similarity between samples or between the detector and samples. Based on the computational process of the negation selection algorithm, matching rules are mainly applied to the generation of detectors during the training phase and the detection of undetermined samples during the testing phase. Matching rules need to be formulated according to the representation of the data and the detector, and are divided into two types: string-based matching rules and real-valued vector-based matching rules. String-based matching rules include r-sequential matching rules, r-chunks matching rules, Hamming distance and its variants, probability-based matching rules, and landscape-affinity matching rules. Real-valued vector-based matching rules can be divided into Minkowski distance, membership function, spatial inclusion matching rules, and bidirectional matching rules.

[0037] In an optional embodiment of the present invention, generating a boundary detector set from the self-sample training set using the nearest neighbor fixed boundary negation selection algorithm includes: dividing the state space into m parts based on the characteristics of the self-samples in the self-sample training set. n Each hypercube is assigned a center point; the properties of each hypercube are determined based on its relationship with the self-sample; and based on the properties of each hypercube, the m... n The outer boundary hypercube is determined from the hypercubes; the boundary detector set is determined based on the outer boundary hypercubes.

[0038] As the core of the negation selection algorithm, the detector generation mechanism determines the algorithm's accuracy and efficiency. How to generate a small number of high-quality detectors in a short time has become crucial to the success of the negation selection algorithm. Similar to the formulation of matching rules, detector generation mechanisms can be divided into two types based on the data and detector representation: string-based generation mechanisms and real-valued vector-based generation mechanisms. String-based generation mechanisms include exhaustive search, linear (greedy) methods, template methods, evolutionary methods, negation databases, and r-variable methods, etc. Real-valued vector-based generation mechanisms include real-valued negation selection algorithms, positive selection algorithms, V-detector algorithms, stepwise methods, adaptive methods, evolutionary methods, and segmentation methods, etc.

[0039] In real-valued vector detector generation mechanisms, the V-detector algorithm is a widely used method. This method generates a set of detectors based on a set of self-samples, and then tries to cover all non-self regions as much as possible. The detector is defined by two parameters: a center point and a radius. The center point is an n-dimensional vector corresponding to a point in the non-self space. The detector is generated using a random method: a center point x is randomly determined in the non-self space, and the distance d between the center point x and the nearest self-sample is determined. A detector with center point x and radius (d-rs) is then generated, where rs is the radius of the nearest self-sample. During the testing phase, if the distance between the sample to be tested and the center point is less than the detector radius, it does not match that detector. If the sample to be tested matches any detector, it is determined to be a non-self sample; otherwise, it is determined to be a self sample.

[0040] The detector is crucial to the negative selection algorithm, and its generation mechanism is the core technology of the algorithm. The algorithm can be roughly divided into a training phase (detector generation phase) and a testing phase (detection phase), as shown in the flowchart below. Figure 2 As shown. Figure 2 A schematic flowchart of the nearest neighbor fixed boundary negation selection algorithm according to an embodiment of the present invention is shown.

[0041] Figure 3 A schematic diagram of the detector structure according to the prior art is shown, such as Figure 3 As shown in the figure, a comparison diagram of a fixed-radius detector and a variable-radius detector is presented. The black area represents the detection blind zone, the square-filled area represents the self-space composed of the self-sample, and the white area represents the generated detector space. The detector's function is to distinguish between self-samples and non-self-samples; complete coverage of the non-self-space is not the only way to achieve this. To reduce the number of detectors, this embodiment generates a layer of detectors around the self-space, identifying the sample to be tested based on the orientation information of the detectors and the self-space. Figure 4A schematic diagram of the structure of a boundary detector according to an embodiment of the present invention is shown, such as... Figure 4 As shown. In this embodiment, the detector layer generated around the self-space is called the boundary detector, which is also determined by two parameters: the center point and the radius.

[0042] Similar to traditional negative selection algorithms, generating a suitable detector is the core problem. First, the state space is uniformly divided into a finite number of hypercubes. Then, the boundary detector is generated based on the information from the hypercubes. To describe the specific process, three definitions are given.

[0043] Definition (1) – Non-empty hypercube When a hypercube h i A hypercube is a non-empty hypercube if it is covered by its own sample or has its own sample inside it, denoted as . h i =Θ; otherwise, it is an empty hypercube, denoted as Θ. h i =О.

[0044] The criterion for determining whether a hypercube is empty is the identification radius δ, which is defined as follows.

[0045] Definition (2) — Recognition radius δ δ is a preset value used to determine whether a hypercube is empty. When a hypercube... h i center point c i If the distance to the nearest self-sample center is greater than δ, then h i =О, otherwise h i =Θ.

[0046] Definition (3) – Outer Boundary Hypercube If a hypercube is empty and its neighboring hypercubes contain at least one non-empty hypercube, then the hypercube is called an outer boundary hypercube.

[0047] In one optional embodiment of the present invention, determining the boundary detector set based on the outer boundary hypercube includes: taking the center of the outer boundary hypercube as the center of the boundary detector, determining the radius of the boundary detector based on the minimum distance to the self-sample, and obtaining the boundary detector corresponding to the outer boundary hypercube; a plurality of boundary detectors constitute the boundary detector set.

[0048] Figure 5 A schematic diagram illustrating the generation of a boundary detector in two-dimensional space according to an embodiment of the present invention is shown, such as... Figure 5As shown. All data were normalized, and the state space was uniformly divided into a 30×30 square grid. The boundary detector was generated based on the outer boundary hypercube, with its center at the center of the outer boundary hypercube. Furthermore, the detector's radius depends on its distance from its nearest self-sample. Therefore, the detector's radius is variable and depends only on the training samples.

[0049] In one optional embodiment of the present invention, determining whether the device status data is healthy based on the relationship between the sample to be tested and the boundary detector includes: determining that the sample to be tested is not in the detection blind zone of the boundary detector; when the sample to be tested is within the boundary detector, the sample to be tested is determined to be a non-self sample, and the device status data is abnormal; when the sample to be tested is not within the boundary detector, if the sample to be tested and a self sample in the self sample training set are located on the same side of the boundary detector, the sample to be tested is determined to be a self sample, and the device status data is normal; when the sample to be tested is not within the boundary detector, if the sample to be tested and a self sample in the self sample training set are located on opposite sides of the boundary detector, the sample to be tested is determined to be a non-self sample, and the device status data is abnormal.

[0050] In some embodiments provided by this invention, if the sample to be tested is in the detection blind zone of the boundary detector, the sample to be tested is determined to be either a self-sample or a non-self-sample by using the nearest sample to the sample to be tested. Figure 5 It can be seen that there are still detection blind zones in both the boundary detector and the self-sample. How to identify the test sample in the detection blind zone is an important factor affecting the accuracy of the algorithm. Therefore, a new matching rule is formulated for the test sample in the detection blind zone. The state of the test sample is determined by determining the attributes of the nearest neighbor sample. The new matching rule is shown in Equation (1).

[0051] (1) In the formula, S represents the self-space; N represents the non-self-space. ds min =min{ d s ij -r j , j =1, 2,…, J}; dd min =min{ dd ik -rab k , k =1, 2,…, K}; dsij For test samples t i With self-sample s j The distance between them; r j The radius of the self-sample; dd ik For test samples t i With boundary detector d k The distance between the center points; rab k Boundary detector d k The radius.

[0052] Figure 6 The schematic diagram illustrates the structure of a device health status assessment system under a fault-free sample according to an embodiment of the present invention, as shown below. Figure 6 As shown. In this embodiment, a device health status assessment system under fault-free sample conditions is also provided. The system includes: a training set generation module, used to construct a self-sample training set based on historical data samples of device status as health status; a detector generation module, used to generate a boundary detector set from the self-sample training set according to the nearest neighbor fixed boundary negation selection algorithm, wherein the boundary detector set includes a plurality of boundary detectors; and a status assessment module, used to take the device status data to be assessed as the test sample, and determine whether the device status data is healthy based on the relationship between the test sample and the boundary detector set.

[0053] In other embodiments, constructing a self-sample training set based on historical data samples where the device health status is normal includes: determining the data representation method of the historical data samples; determining the representation method of the boundary detector based on the data representation method of the historical data samples; and determining the matching rules between the historical data samples and the boundary detector.

[0054] In other embodiments, generating a boundary detector set from the self-sample training set using a nearest neighbor fixed boundary negation selection algorithm includes: The state space is divided into m parts based on the characteristics of the self-samples in the self-sample training set. n Each hypercube is assigned a center point; the properties of each hypercube are determined based on its relationship with the self-sample; and based on the properties of each hypercube, the m... n The outer boundary hypercube is determined from the hypercubes; the boundary detector set is determined based on the outer boundary hypercubes.

[0055] In other embodiments, the properties of each hypercube are determined based on the relationship between each hypercube and the self-sample, including: when a hypercube is covered by the self-sample, or the self-sample is inside the hypercube, the hypercube is determined to be a non-empty hypercube; otherwise, the hypercube is determined to be an empty hypercube.

[0056] In other embodiments, based on the properties of each hypercube, from the m n Determining the outer boundary hypercube among a hypercube includes: if a hypercube is an empty hypercube and at least one of its neighboring hypercubes is a non-empty hypercube, then the hypercube is an outer boundary hypercube.

[0057] In other embodiments, determining the boundary detector set based on the outer boundary hypercube includes: using the center of the outer boundary hypercube as the center of the boundary detector, determining the radius of the boundary detector based on the minimum distance to the self-sample, and obtaining the boundary detector corresponding to the outer boundary hypercube; a plurality of boundary detectors constitute the boundary detector set.

[0058] In other embodiments, the device status data is determined to be normal based on the relationship between the sample to be tested and the boundary detector, including: determining that the sample to be tested is not in the detection blind zone of the boundary detector; When the sample to be tested is within the boundary detector, the sample to be tested is determined to be a non-self sample, and the device status data is abnormal. When the sample to be tested is not within the boundary detector, if the sample to be tested and the self sample in the self sample training set are located on the same side of the boundary detector, then the sample to be tested is determined to be a self sample, and the device status data is normal. When the sample to be tested is not within the boundary detector, if the sample to be tested and the self sample in the self sample training set are located on opposite sides of the boundary detector, then the sample to be tested is determined to be a non-self sample, and the device status data is abnormal.

[0059] In other implementations, if the sample to be tested is in the detection blind zone of the boundary detector, the sample to be tested is determined to be either a self-sample or a non-self-sample by the nearest sample to the sample to be tested.

[0060] The specific limitations of each functional module in the above-mentioned equipment health status assessment system under fault-free samples can be found in the limitations of the equipment health status assessment method under fault-free samples mentioned above, and will not be repeated here. Each module in the above-mentioned device can be implemented entirely or partially through software, hardware, or a combination thereof. Each module can be embedded in the processor of the computer device in hardware form or independent of it, or it can be stored in the memory of the computer device in software form, so that the processor can call and execute the operations corresponding to each module.

[0061] In one embodiment of the present invention, an electronic device is also provided, comprising: at least one processor; and a memory connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the at least one processor implements the aforementioned device health status assessment method under fault-free sample conditions by executing the instructions stored in the memory. The control module or processor here has numerical calculation and logical operation functions, and at least has a central processing unit (CPU) with data processing capabilities, random access memory (RAM), read-only memory (ROM), multiple I / O ports, and an interrupt system. The processor contains a kernel, which retrieves corresponding program units from the memory. One or more kernels can be configured, and the aforementioned method can be implemented by adjusting kernel parameters. The memory may include non-permanent memory in computer-readable media, random access memory (RAM), and / or non-volatile memory, such as read-only memory (ROM) or flash RAM, and the memory includes at least one memory chip.

[0062] In one embodiment of the present invention, a machine-readable storage medium is also provided, on which instructions are stored, which, when executed by a processor, cause the processor to be configured to perform the aforementioned device health status assessment method under a fault-free sample.

[0063] In one embodiment of the present invention, a computer program product is provided, including a computer program that, when executed by a processor, implements the above-described method for assessing the health status of equipment under fault-free conditions.

[0064] To verify the effectiveness of the embodiments provided in this invention, the Iris dataset is used for validation. This invention utilizes the Iris dataset from the UCI standard database to verify the effectiveness of the proposed NFB-NSA method. The computer environment used for implementation was a 3.60 GHz Intel(R) Core(TM) i7-9700U CPU, 16GB of RAM, Windows 10, 64-bit operating system, and the relevant algorithm was implemented using Python 3.6.

[0065] The Iris dataset, also known as the iris flower dataset, is commonly used for multivariate classification. It contains three classes of iris flower samples: Setosa, Versicolor, and Viginica. Each class is distinguished by four attributes: sepal length, sepal width, petal length, and petal width. Each iris flower class contains 50 data points. Therefore, the Iris dataset is a 150×4 matrix.

[0066] For ease of calculation, the data is first normalized before application. The normalization formula is shown in equation (2). After normalization, the Iris dataset can be projected into a two-dimensional space.

[0067] (2) In the formula, min and max are the minimum and maximum values ​​of the sample data, respectively; x It is the raw data. x m It is the normalized data.

[0068] To characterize the performance of the nearest neighbor fixed boundary negation selection algorithm and to compare it with other high-recognition-rate methods, the Iris dataset was analyzed. Three evaluation metrics were defined: detection rate (DR), false alarm rate (FA), and detection accuracy (DPR), with the specific calculation formulas as follows.

[0069] (3) (4) DPR = DR / (DR + FA) (5) In the formula, TP and FN These represent the number of self samples that were correctly identified and the number of samples that were incorrectly classified as non-self samples, respectively. FP and TN These represent the number of non-self samples that were correctly identified and the number of samples that were incorrectly classified as self samples, respectively.

[0070] Three evaluation metrics characterize the algorithm's performance from three aspects. The detection rate characterizes the accuracy of the detector generated by the algorithm in recognizing self-samples; the false alarm rate characterizes the error rate of the detector generated by the algorithm in recognizing non-self-samples; and the detection accuracy comprehensively considers the algorithm's detection performance on both self-samples and non-self-samples.

[0071] The proposed nearest neighbor fixed boundary negation selection algorithm was used to analyze the Iris dataset. Similar to other comparative methods, 100% of the Setosa data, 50% of the Setosa data, 100% of the Versicolor data, 50% of the Versicolor data, 100% of the Viginica data, and 50% of the Viginica data were used as self-samples for 100 repeated calculations. Then, the detection rate, false alarm rate, and detection accuracy of each algorithm were calculated under different self-samples, as shown in Table 1 below.

[0072] Table 1 shows that when Setosa and Versicolor are used as self-samples, NFB-NSA achieves the highest detection rate, false alarm rate, and detection accuracy, with the smallest standard deviations in both detection rate and false alarm rate, indicating that the algorithm's computation remains relatively stable. When Viginica is used as a self-sample, while NFB-NSA's detection rate is not the highest, it is close to the highest, and it has the lowest false alarm rate and the highest detection accuracy. In summary, the nearest neighbor fixed boundary negation selection algorithm exhibits better detection performance and maintains relatively stable computational results.

[0073] During the calculation process, the computation time and the number of detectors generated by each algorithm were statistically analyzed. Compared with other algorithms, the nearest neighbor fixed boundary negation selection algorithm proposed in this paper generates fewer detectors while maintaining accurate detection results, which greatly reduces the amount of computation and thus takes less time.

[0074] Table 1 Comparison of Iris dataset recognition results

[0075] To enable those skilled in the art to better understand the embodiments of the present invention, field data from an electrically driven centrifugal compressor unit at a gas transmission station are used to evaluate the effectiveness of the method. Figure 7 A schematic diagram of the structure of a centrifugal compressor unit according to an embodiment of the present invention is shown, such as... Figure 7 As shown. The key subsystems of the compressor unit include the centrifugal compressor, the electric motor, and other major auxiliary systems, such as the sealing system, gas cooling system, and lubrication system. The actual operating parameters of the compressor need to be adjusted according to the supply and demand of upstream and downstream, especially with significant differences between summer and winter. Therefore, to improve the comprehensiveness and accuracy of the early warning, 2400 sets of normal operating data were selected from the annual operating data of the compressor unit for analysis. The 2400 sets of data were evenly distributed across 12 months, i.e., 200 sets of data were selected from each month, and all data were randomly selected. In addition, to verify the algorithm's effectiveness in identifying abnormal states, 10 sets of operating data indicating abnormal unit conditions were collected.

[0076] To obtain more accurate early warning results, a compressor unit early warning index system containing seven monitoring parameters was established for calculation (exhaust pressure, motor speed, compressor shaft vibration displacement (#2), inlet pressure, air to third-stage seal pressure (#4), motor drive end vibration displacement (X), and lubricating oil head temperature). Early warning calculations were performed under six different training sample sizes: 120, 240, 360, 480, 600, and 720 data sets. Similar to the previously selected 2400 data sets, all training sample data were uniformly taken from 12 months and were randomly selected. Then, the proposed nearest neighbor fixed boundary negation selection algorithm was used for unit early warning under the six training samples. To ensure the stability of the calculation results, the calculation was repeated 10 times for each training sample. The average detection rate and average false alarm rate were used to characterize the algorithm's early warning performance. The specific calculation results of the indicators are as follows: Figure 8 As shown. Figure 8 A schematic diagram illustrating the comparison of early warning results under different training sample sizes according to an embodiment of the present invention is provided. The diagram shows that only when the number of training samples is 120 did a false alarm occur; no false alarms occurred at other training sample sizes. The lowest average detection rate occurred when the number of training samples was 120, at 85.18%; the highest average detection rate occurred when the number of training samples was 480, at 97.03%. Overall, as the training sample size increases from 120 to 480, the average detection rate gradually increases, and the rate of increase shows a decreasing trend; when the training sample size increases from 480 to 720, the average detection rate does not change significantly and remains basically stable.

[0077] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0078] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0079] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0080] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0081] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0082] Memory may include non-persistent memory in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, like read-only memory (ROM) or flash memory (flash RAM). Memory is an example of computer-readable media.

[0083] Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can store information using any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0084] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0085] The above are merely embodiments of this application and are not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.

Claims

1. A method for assessing equipment health status under fault-free conditions, characterized in that, The equipment is a compressor unit, and the method includes: A training set of self-samples is constructed based on historical data samples of the equipment status data for the health status; the equipment status data includes: exhaust pressure, motor speed, vibration displacement of the compressor shaft, intake pressure, air to the third-level seal pressure, vibration displacement of the motor drive end, and lubricating oil head temperature; Generating a boundary detector set from the self-sample training set using the nearest neighbor fixed boundary negation selection algorithm includes: dividing the state space into m parts based on the characteristics of the self-samples in the self-sample training set. n Each hypercube is assigned a center point; the properties of each hypercube are determined based on its relationship with the self-sample; and based on the properties of each hypercube, the m... n An outer boundary hypercube is determined from a set of hypercubes; a set of boundary detectors is determined based on the outer boundary hypercubes; the set of boundary detectors includes a plurality of boundary detectors; The device status data to be evaluated is used as a test sample. Based on the relationship between the test sample and the boundary detector set, it is determined whether the device status data to be evaluated is healthy. Based on the relationship between each hypercube and the self-sample, the properties of each hypercube are determined, including: When a hypercube is covered by the self-sample, or the self-sample is inside the hypercube, the hypercube is determined to be a non-empty hypercube. Otherwise, the hypercube is determined to be an empty hypercube. Based on the properties of each hypercube, from the m n The outer boundary hypercube is determined from among the hypercubes, including: If a hypercube is an empty hypercube and at least one of its neighboring hypercubes is a non-empty hypercube, then the hypercube is an outer boundary hypercube. The boundary detector set is determined based on the outer boundary hypercube, including: Using the center of the outer boundary hypercube as the center of the boundary detector, the radius of the boundary detector is determined based on the minimum distance to the self-sample, thus obtaining the boundary detector corresponding to the outer boundary hypercube; The boundary detector set consists of several boundary detectors; If the sample to be tested is in the detection blind zone of the boundary detector, then the sample to be tested is determined to be either a self-sample or a non-self-sample, and thus the device status data corresponding to the sample to be tested is determined to be healthy or unhealthy, specifically: ； In the formula, S represents the self-space; N represents the non-self-space. ds min =min{ d s ij -r j , j =1, 2,…, J }; dd min =min{ dd ik - rab k , k =1, 2,…, K }; ds ij For test samples t i With self-sample s j The distance between them; r j The radius of the self-sample; dd ik For test samples t i With boundary detector d k The distance between the center points; rab k Boundary detector d k The radius.

2. The method according to claim 1, characterized in that, Based on historical data samples of device status, a self-sample training set is constructed, including: Determine the data representation method of the historical data sample; The representation of the boundary detector is determined based on the data representation of the historical data samples; and Determine the matching rules between the historical data samples and the boundary detector.

3. The method according to claim 1, characterized in that, Determining the health status of the device to be evaluated based on the relationship between the sample to be tested and the boundary detector includes: Determine that the sample to be tested is not in the detection blind zone of the boundary detector; When the sample to be tested is within the boundary detector, the sample to be tested is determined to be a non-self sample, and the device status data to be evaluated is unhealthy. When the sample to be tested is not within the boundary detector, if the sample to be tested and the self sample in the self sample training set are located on the same side of the boundary detector, then the sample to be tested is determined to be a self sample, and the device status data to be evaluated is healthy. When the sample to be tested is not within the boundary detector, if the sample to be tested and the self sample in the self sample training set are located on opposite sides of the boundary detector, then the sample to be tested is determined to be a non-self sample, and the device status data to be evaluated is unhealthy.

4. A system for assessing equipment health status under fault-free conditions, characterized in that, The system is used to implement the equipment health status assessment method under a fault-free sample as described in any one of claims 1 to 3, the system comprising: The training set generation module is used to construct a self-sample training set based on historical data samples of device health status using device status data. A detector generation module is used to generate a boundary detector set from the self-sample training set using a nearest neighbor fixed boundary negation selection algorithm, wherein the boundary detector set includes a plurality of boundary detectors; and The status assessment module is used to take the status data of the device to be assessed as a test sample, and determine whether the status data of the device is healthy based on the relationship between the test sample and the boundary detector set.

5. The system according to claim 4, characterized in that, Based on historical data samples of device status, a self-sample training set is constructed, including: Determine the data representation method of the historical data sample; The representation of the boundary detector is determined based on the data representation of the historical data samples; and Determine the matching rules between the historical data samples and the boundary detector.

6. The system according to claim 5, characterized in that, The self-sample training set is used to generate a boundary detector set according to the nearest neighbor fixed boundary negation selection algorithm, including: Based on the characteristics of the self-samples in the self-sample training set, the state space is divided into m... n Given a hypercube, obtain the center point of each hypercube; The properties of each hypercube are determined based on its relationship with the self-sample; Based on the properties of each hypercube, from the m n Determine the outer boundary hypercube from among the hypercubes; The boundary detector set is determined based on the outer boundary hypercube.

7. The system according to claim 6, characterized in that, Based on the relationship between each hypercube and the self-sample, the properties of each hypercube are determined, including: When a hypercube is covered by the self-sample, or the self-sample is inside the hypercube, the hypercube is determined to be a non-empty hypercube. Otherwise, the hypercube is determined to be an empty hypercube.

8. The system according to claim 7, characterized in that, Based on the properties of each hypercube, from the m n The outer boundary hypercube is determined from among the hypercubes, including: A hypercube is an outer boundary hypercube if it is an empty hypercube and at least one of its neighboring hypercubes is a non-empty hypercube.

9. The system according to claim 8, characterized in that, The boundary detector set is determined based on the outer boundary hypercube, including: Using the center of the outer boundary hypercube as the center of the boundary detector, the radius of the boundary detector is determined based on the minimum distance to the self-sample, thus obtaining the boundary detector corresponding to the outer boundary hypercube; Several boundary detectors constitute the boundary detector set.

10. The system according to claim 4, characterized in that, Determining the health status of the device based on the relationship between the sample to be tested and the boundary detector includes: Determine that the sample to be tested is not in the detection blind zone of the boundary detector; When the sample to be tested is within the boundary detector, the sample to be tested is determined to be a non-self sample, and the device status data is unhealthy. When the sample to be tested is not within the boundary detector, if the sample to be tested and the self sample in the self sample training set are located on the same side of the boundary detector, then the sample to be tested is determined to be a self sample, and the device status data is healthy. When the sample to be tested is not within the boundary detector, if the sample to be tested and the self sample in the self sample training set are located on opposite sides of the boundary detector, then the sample to be tested is determined to be a non-self sample, and the device status data is unhealthy.

11. The system according to claim 10, characterized in that, If the sample to be tested is in the detection blind zone of the boundary detector, the sample to be tested is determined to be either a self-sample or a non-self-sample, and the device status data corresponding to the sample to be tested is determined to be healthy or unhealthy.

12. An electronic device, characterized in that, include: At least one processor; A memory connected to the at least one processor; The memory stores instructions that can be executed by the at least one processor, and the at least one processor implements the equipment health status assessment method under a fault-free sample as described in any one of claims 1 to 3 by executing the instructions stored in the memory.

13. A machine-readable storage medium, characterized in that, The machine-readable storage medium stores instructions that, when executed by a processor, cause the processor to perform the device health status assessment method under a fault-free sample as described in any one of claims 1 to 3.

Images