Familial defect identification method of hydropower station industrial control system
By dividing the hydropower station's industrial control system into functional units and constructing a feature vector set, and using a combination weighting method and Euclidean distance calculation, the problem of early identification and accurate location of family-related defects was solved, thereby improving the operational reliability and fault handling efficiency of hydropower station equipment.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-01-26
- Publication Date
- 2026-05-01
AI Technical Summary
Existing technologies struggle to accurately identify family-related defects in hydropower station control systems, leading to frequent unplanned equipment outages that affect the normal operation of the power system. Furthermore, they lack early warning capabilities and have insufficient accuracy in identification.
The industrial control system is divided into multiple units according to function, a dual-identification feature vector set is constructed, the feature vector weights are calculated by the combined weighting method, and the Euclidean distance and ROC curve are dynamically adjusted to achieve early identification and accurate location of family defects.
It enables real-time monitoring and accurate identification of family-related defects in industrial control systems during operation, reduces unplanned downtime, improves equipment reliability, and provides efficient defect root cause tracing and alarm guidance.
Smart Images
Figure QLYQS_1 
Figure QLYQS_2 
Figure QLYQS_3
Abstract
Description
Technical Field
[0001] This invention belongs to the field of automatic control technology for hydropower stations, and specifically relates to a method for identifying family-related defects in hydropower station industrial control systems. Background Technology
[0002] In the current power system, the installed capacity of intermittent renewable energy sources such as wind and solar power is constantly expanding. In order to ensure the integration effect of intermittent renewable energy sources with traditional power systems, it is necessary to carry out large-scale digital and intelligent transformation of existing hydropower stations to meet the needs of such new power systems for safe, stable operation and efficient management.
[0003] To ensure the safe and stable operation of hydropower units and the power grid, in-depth fault diagnosis research on various industrial control systems is imperative. Among these, family defects refer to defects caused by common factors in power equipment, such as design, materials, or manufacturing processes. This means that due to shared design schemes, component batches, supplier sources, or operating environments, equipment units of the same model or batch exhibit the same or similar inherent defects. These are major defects that lead to concentrated quality risks in equipment and components, seriously affecting the safe operation of transmission channels. Furthermore, traditional defect detection methods often rely on the numerical fluctuations of a single parameter over a certain period for assessment, calculation, and statistics. This makes it difficult to accurately count family defects with multi-attribute correlations on various problematic equipment, resulting in inaccurate and incomplete defect analysis conclusions. Consequently, it becomes difficult to identify the true causes of power system operational problems and effectively eliminate operational faults.
[0004] Currently, although methods for detecting family-related defects have been improved, involving the detection and judgment of multiple related data, these methods often only detect and trigger fault alarms when the related target parameters detected by a certain industrial control system exceed the set equipment parameter range, or when a component malfunctions due to damage. However, these methods are often applied to post-event judgment, lacking the ability to identify family-related defects early, leading to more unplanned shutdowns and maintenance of electromechanical equipment, affecting the normal operation of the power system. Furthermore, because the early signs of some family-related defects are not obvious, some defects are difficult for staff to detect or for testing instruments to accurately identify, increasing the difficulty of family-related defect identification and affecting its accuracy. Therefore, there is an urgent need for a family-related defect identification method for hydropower station industrial control systems that can solve the above problems. Summary of the Invention
[0005] The purpose of this invention is to address the aforementioned shortcomings by providing a method for identifying family-related defects in hydropower station control systems. This method aims to solve the problems of current family-related defect identification methods lacking early warning capabilities, leading to numerous unplanned equipment outages that affect the normal operation of the power system, and also lacking sufficient accuracy. To achieve the above objective, this invention provides the following technical solution:
[0006] A method for identifying family-related defects in a hydropower station's industrial control system includes the following steps: S1: Divide the hydropower station's industrial control system into multiple functional units and encode them digitally; S2: Construct a dual-identification feature vector set for each functional unit, and sort and normalize each feature vector in the dual-identification feature vector set; wherein, the feature vectors in the dual-identification feature vector set are arranged in the following order: defect severity S, defect occurrence frequency F, component batch consistency B, operating environment sensitivity E and design scheme reusability D, component supplier concentration V, component life loss rate L and component defect type consistency T. S3: The combined weighting method is used to calculate and determine the combined weight w of each feature vector in each functional unit; Furthermore, the calculated weights w are sorted according to the eigenvector order of each functional unit, i.e., by w. S w F w B w E w D w V w L w T Arrange them in order to facilitate subsequent analysis of common defects in the industrial control system; S4: Based on the feature vector data and weights of normal functional units without familial defects, calculate the cluster center vector in their healthy state as the identification benchmark vector. S5: Calculate the first relative Euclidean distance d1 between each feature vector in the target functional unit and its corresponding feature vector in the identification reference vector, and the second relative Euclidean distance d' between each feature vector in the target functional unit and its corresponding feature vector in each functional unit. ia Then, based on d1 and d' ia Calculate the familial defect difference d2 between any two adjacent functional units; The formula for calculating the first relative Euclidean distance d1 is: ; In the formula, k is the index of the eigenvector, and w k x is the combined weight of the k-th eigenvector in the target functional unit. kb is the normalized value of the k-th eigenvector of the target functional unit. 0k To identify the normalized value of the k-th eigenvector in the reference vector; Second relative Euclidean distance d' ia The formula for calculating the familial defect difference d2 is: ; ; In the formula, i is the sequence number of the target functional unit, a is the sequence number of each functional unit in the industrial control system, and i ≠ a; j is the sequence number of the feature vector in the functional unit; N is the number of functional units in the industrial control system; d' ia It is the second relative Euclidean distance between any two functional units; In addition, in the formula This is partly to eliminate duplicate coefficients and ensure that all pairwise combinations are calculated only once; S6: Based on the historical feature vector data of functional units with familial defects and the historical feature vector data of normal functional units, construct the ROC working curve and make dynamic adjustments to obtain and correct the dynamic distance adjustment threshold y0 corresponding to different functional units. S7: Based on the first relative Euclidean distance d1 and the second relative Euclidean distance d' of each functional unit ia The familial defect difference degree d2 and the corresponding dynamic distance adjustment threshold y0 are used to determine the familial defect status of each functional unit. If d1 > y0 of the target functional unit, the functional unit is determined to have a single family defect; if no less than 3 functional units simultaneously satisfy d1 > y0 and d2 is less than the upper limit of the preset second Euclidean distance d2, the industrial control system is determined to have a cluster family defect. S8: For industrial control systems with clustered or family-like defects, trace the root cause of the defects based on the commonalities of the defects.
[0007] Furthermore, in step S1, the functional units of the industrial control system include at least a power supply unit, a PLC controller unit, an execution drive unit, a key sensing element unit, and a communication interface unit.
[0008] Further, in step S2, the dual identification feature vector set is divided into a unit dimension vector subset and a component dimension vector subset; wherein, the defect severity S, defect occurrence frequency F, component batch consistency B, operating environment sensitivity E, and design scheme reusability D belong to the unit dimension vector subset; and the component supplier concentration V, component life loss rate L, and component defect type consistency T belong to the component dimension vector subset.
[0009] Furthermore, in step S2, the construction of each feature vector is as follows: The severity of defects (S) is divided into five levels from low to high based on the severity of the consequences: no defects, false alarms, parameter drift, logic errors, and program crashes. The defect frequency F is calculated based on the number of defects occurring every 30 days, a. ; Component batch consistency B is calculated based on the proportion of components in the same batch to the total number of components; The sensitivity of the operating environment E is calibrated based on the increase in defect incidence when the operating environment temperature is >45℃, and the maximum value is taken when the defect incidence increases to two times or more. The reusability D of the design scheme is calculated based on the reuse ratio of the PLC control program. Supplier concentration V is calculated as the proportion of the number of core components within a functional unit that come from the same supplier to the total number of core components. The life loss rate L is calculated as the ratio of the remaining life of the component to its design life. Defect type consistency T is calculated based on the proportion of the number of defects of the same type in a functional unit to the total number of defects within a preset statistical period.
[0010] Furthermore, the normalization method used in step S2 is specifically the maximum-minimum method.
[0011] Furthermore, in step S2, within a single functional unit, each feature vector is normalized to form a dual-identification feature vector set, as shown below: ; Where X1 is the set of feature vectors of the first functional unit; x 11 The first dimension is the feature vector; Then, let X be the set containing the dual-identification feature vector sets of all functional units, and X is expressed as: .
[0012] Furthermore, the combined weighting method in step S3 specifically refers to: For each feature vector in each functional unit, the subjective and objective weights are calculated using the analytic hierarchy process (AHP) and the entropy weighting method, respectively. Then, the final combined weight of the target feature vector is obtained through a combination operation, and its calculation formula is as follows: W k =Subjective weight × 0.6 + Objective weight × 0.4.
[0013] Furthermore, in step S6, the specific method for dynamically correcting the ROC curve is as follows: whenever 50 new sets of feature vector data with familial defects are added, the ROC curve is recalculated based on the new data and the historical feature vector data of the normal functional unit, and the corrected dynamic distance adjustment threshold y0 is obtained.
[0014] Furthermore, in step S7, the preset upper limit of the second Euclidean distance threshold is 0.3.
[0015] Furthermore, in step S8, the method for tracing the root cause of the defect is specifically as follows: S81: Preset common defect weight threshold w 缺 In each functional unit exhibiting clustered or family-like defects, extract the combined weight value w. k Higher than w 缺 The eigenvectors are analyzed, and the root cause of familial defects is determined based on the combination of each eigenvector and its subset. S82: Based on the judgment result, send out different alarm messages.
[0016] The beneficial effects of this invention are: 1. This invention quantifies the various defect feature vectors that may exist in equipment and components into standardized data, and uses calculation and judgment to achieve early detection of anomalies in family defects. This not only expands the coverage of family defect identification but also significantly improves the accuracy of family defect identification. Furthermore, this method can monitor the changes in family defect-related feature vectors in real time during the operation of the industrial control system. It can accurately detect and identify the types of family defects and issue timely alarms while detecting anomalies, achieving the effect of advance judgment. This allows staff to intervene in a timely manner, reducing the number of unplanned shutdowns and maintenance of electromechanical equipment and improving the reliability of power plant electromechanical equipment operation.
[0017] 2. This invention traces the root causes of identified clustered and family-related defects based on their common high-weight characteristics, linking specific aspects such as environment, defect type, and procurement. This enables staff to quickly determine the type of family-related defect based on different alarm conditions, providing efficient and accurate guidance for subsequent maintenance and improvement strategy development. Detailed Implementation
[0018] In the description of this invention, it should be understood that the terms "center," "longitudinal," "lateral," "length," "width," "thickness," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," "outer," "clockwise," "counterclockwise," "axial," "radial," and "circumferential" indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are used only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this invention.
[0019] In the description of this invention, "first feature" and "second feature" may include one or more of the features.
[0020] In the description of this invention, "a plurality of" means two or more.
[0021] In the description of this invention, the first feature being "above" or "below" the second feature may include the first and second features being in direct contact, or it may include the first and second features not being in direct contact but being in contact through another feature between them.
[0022] In the description of this invention, the terms "above," "over," and "on top" for the first feature and the second feature include the first feature being directly above or diagonally above the second feature, or simply indicating that the first feature is at a higher horizontal level than the second feature.
[0023] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "illustrative embodiment," "example," "specific example," and "some examples" indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.
[0024] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments, but the present invention is not limited to the following embodiments.
[0025] Example 1: A method for identifying family-related defects in a hydropower station's industrial control system includes the following steps: S1: Divide the hydropower station's industrial control system into multiple functional units and encode them digitally; S2: Construct a dual-identification feature vector set for each functional unit, and sort and normalize each feature vector in the dual-identification feature vector set; wherein, the feature vectors in the dual-identification feature vector set are arranged in the following order: defect severity S, defect occurrence frequency F, component batch consistency B, operating environment sensitivity E and design scheme reusability D, component supplier concentration V, component life loss rate L and component defect type consistency T. S3: The combined weighting method is used to calculate and determine the combined weight w of each feature vector in each functional unit; Furthermore, the calculated weights w are sorted according to the eigenvector order of each functional unit, i.e., by w. S w F w B w E w D w V w L w T Arrange them in order to facilitate subsequent analysis of common defects in the industrial control system; S4: Based on the feature vector data and weights of normal functional units without familial defects, calculate the cluster center vector in their healthy state as the identification benchmark vector. S5: Calculate the first relative Euclidean distance d1 between each feature vector in the target functional unit and its corresponding feature vector in the identification reference vector, and the second relative Euclidean distance d' between each feature vector in the target functional unit and its corresponding feature vector in each functional unit. ia Then, based on d1 and d' ia Calculate the familial defect difference d2 between any two adjacent functional units; The formula for calculating the first relative Euclidean distance d1 is: ; In the formula, k is the index of the eigenvector, and w k x is the combined weight of the k-th eigenvector in the target functional unit. k b is the normalized value of the k-th eigenvector of the target functional unit. 0k To identify the normalized value of the k-th eigenvector in the reference vector; Second relative Euclidean distance d' ia The formula for calculating the familial defect difference d2 is: ; ; In the formula, i is the sequence number of the target functional unit, a is the sequence number of each functional unit in the industrial control system, and i ≠ a; j is the sequence number of the feature vector in the functional unit; N is the number of functional units in the industrial control system; d' ia It is the second relative Euclidean distance between any two functional units; In addition, in the formula This is partly to eliminate duplicate coefficients and ensure that all pairwise combinations are calculated only once; S6: Based on the historical feature vector data of functional units with familial defects and the historical feature vector data of normal functional units, construct the ROC working curve and make dynamic adjustments to obtain and correct the dynamic distance adjustment threshold y0 corresponding to different functional units. S7: Based on the first relative Euclidean distance d1 and the second relative Euclidean distance d' of each functional unit ia The familial defect difference degree d2 and the corresponding dynamic distance adjustment threshold y0 are used to determine the familial defect status of each functional unit. If d1 > y0 of the target functional unit, the functional unit is determined to have a single family defect; if no less than 3 functional units simultaneously satisfy d1 > y0 and d2 is less than the upper limit of the preset second Euclidean distance d2, the industrial control system is determined to have a cluster family defect. S8: For industrial control systems with clustered or family-like defects, trace the root cause of the defects based on the commonalities of the defects.
[0026] As described above, this invention first divides a complex industrial control system into multiple functional units and constructs a dual-identification feature vector set that integrates multi-dimensional feature vectors for each unit. This transforms the common factors that may lead to multi-dimensional familial defects into unified and calculable values. Based on the identification benchmark vector, it calculates the first relative Euclidean distance d1, the second relative Euclidean distance d2, and the familial defect difference d2 between any two adjacent functional units for each feature vector in the target functional unit. This quantifies the degree to which the target functional unit deviates from the health standard. Then, through intelligent threshold judgment and cluster analysis, it completes the accurate identification and classification of defects. Finally, it can also locate the root cause of defects by tracing high-weight common features.
[0027] This invention can monitor and accurately identify whether there are family-related defects in the industrial control system during operation, and send out alarms based on the root cause of the defects, guiding maintenance personnel to carry out precise and efficient maintenance, thereby improving the overall reliability of the power plant's electromechanical equipment.
[0028] Example 2: Based on Embodiment 1, in step S1, the functional units of the industrial control system include at least a power supply unit, a PLC controller unit, an execution drive unit, a key sensing element unit, and a communication interface unit.
[0029] In step S2, the dual identification feature vector set is divided into a unit dimension vector subset and a component dimension vector subset; among them, the defect severity S, defect occurrence frequency F, component batch consistency B, operating environment sensitivity E, and design scheme reusability D belong to the unit dimension vector subset; the component supplier concentration V, component life loss rate L, and component defect type consistency T belong to the component dimension vector subset.
[0030] As can be seen from the above, the number of functional units in an industrial control system can be designed according to actual operational requirements. In this embodiment, the number of functional units in the industrial control system is set to five, i.e., N=5. Furthermore, the dual-identification feature vector set is divided into a unit-dimensional vector subset and a component-dimensional vector subset to support subsequent defect tracing analysis.
[0031] The feature vectors of the unit dimension vector subset focus on the overall performance of the functional unit during operation. Defect severity (S) and defect frequency (F) directly reflect the unit's reliability level from its operational history, while component batch consistency (B), operating environment sensitivity (E), and design reuse (D) reflect the inherent risks of the functional unit from the perspective of design, configuration, and deployment environments. The component dimension vector subset, including component supplier concentration (V), component lifespan loss rate (L), and component defect type consistency (T), focuses on the common characteristics of key components within the functional unit in specific stages such as manufacturing, supply chain management, and maintenance.
[0032] Example 3: Based on Example 2, in step S2, the construction methods of each feature vector are as follows: The severity of defects (S) is divided into five levels from low to high based on the severity of the consequences: no defects, false alarms, parameter drift, logic errors, and program crashes. The defect frequency F is calculated based on the number of defects occurring every 30 days, a. ; Component batch consistency B is calculated based on the proportion of components in the same batch to the total number of components; The sensitivity of the operating environment E is calibrated based on the increase in defect incidence when the operating environment temperature is >45℃, and the maximum value is taken when the defect incidence increases to two times or more. The reusability D of the design scheme is calculated based on the reuse ratio of the PLC control program. Supplier concentration V is calculated as the proportion of the number of core components within a functional unit that come from the same supplier to the total number of core components. The life loss rate L is calculated as the ratio of the remaining life of the component to its design life. Defect type consistency T is calculated based on the proportion of the number of defects of the same type in a functional unit to the total number of defects within a preset statistical period.
[0033] The normalization method used in step S2 is specifically the maximum-minimum method.
[0034] In step S2, within a single functional unit, each feature vector is normalized to form a dual-identification feature vector set, as shown below: ; Where X1 is the set of feature vectors of the first functional unit; x 11 The first dimension is the feature vector; Then, let X be the set containing the dual-identification feature vector sets of all functional units, and X is expressed as: .
[0035] Furthermore, the combined weighting method in step S3 specifically refers to: For each feature vector in each functional unit, the subjective and objective weights are calculated using the analytic hierarchy process (AHP) and the entropy weighting method, respectively. Then, the final combined weight of the target feature vector is obtained through a combination operation, and its calculation formula is as follows: W k =Subjective weight × 0.6 + Objective weight × 0.4.
[0036] As described above, this invention transforms the multidimensional and vaguely defined vector features in the dual-identification feature vector set into standardized numerical values, providing a unified numerical basis for subsequent calculation of their respective weights w and the determination of familial defects. Specifically, all feature vectors are uniformly quantized into values within the [0,1] interval. The method of combining the analytic hierarchy process (AHP) and the entropy weight method to calculate subjective and objective weights separately is existing technology and will not be elaborated upon here. Finally, the subjective and objective weights are weighted again to obtain the combined weight w of a single feature vector.
[0037] Example 4: Based on Example 3, in step S6, the specific method for dynamically correcting the ROC curve is as follows: whenever 50 new sets of feature vector data with familial defects are added, the ROC curve is recalculated based on the new data and the historical feature vector data of the normal functional unit, and the corrected dynamic distance adjustment threshold y0 is obtained.
[0038] In step S7, the preset upper limit of the second Euclidean distance threshold is 0.3.
[0039] As can be seen from the above, the present invention calculates and dynamically adjusts the ROC curve to optimize the dynamic distance adjustment threshold y0 in real time during the operation of the industrial control system, thereby avoiding misjudgment due to changes in the equipment operating status caused by a fixed threshold and ensuring the accuracy of identifying family defects.
[0040] Example 5: Based on Example 4, in step S8, the method for tracing the root cause of the defect is specifically as follows: S81: Preset common defect weight threshold w 缺 In each functional unit exhibiting clustered or family-like defects, extract the combined weight value w. k Higher than w 缺 The eigenvectors are analyzed, and the root cause of familial defects is determined based on the combination of each eigenvector and its subset. S82: Based on the judgment result, send out different alarm messages.
[0041] As can be seen from the above, the present invention uses a preset common defect weight threshold w 缺 This is compared with the weights of the vector features of each functional unit that have been calculated and obtained above, in order to help staff trace the root causes of potential defects in the industrial control unit.
[0042] Specifically, when the system determines that a clustered or family-like defect exists, it will automatically analyze all functional units with clustered or family-like defects. During the comparative analysis, the analysis order is based on the order of the weights w of each vector feature combination. S w F w B w E w D w V w L w T The comparisons and extractions are performed sequentially. The combined weight value w is then used in each functional unit. k Higher than w 缺 The key feature vectors are those that reflect the direction of defects that have the greatest impact and the most significant commonality in the defect cluster. Then, based on the combinations of these key feature vectors and their respective subsets, classification judgments can be made according to preset conditions. For example, when the key feature vector is a combination of component batch consistency B and supplier concentration V, it may be a batch nature problem of a single supplier, and the root cause of the defect can be determined to be the procurement and supply chain links. When the key feature vector is a combination of design reuse degree D and defect type consistency T, it indicates that there is an inherent defect in the reuse design, and the root cause of the defect is determined to be the design and engineering process. When the key feature vector is a combination of operating environment sensitivity E and life loss rate L, it indicates that the industrial control system is experiencing accelerated wear and tear due to an inappropriate operating environment. Therefore, the root cause of its defects is determined to be the operation and maintenance and external environment.
[0043] Finally, based on the assessment results, targeted alarm messages with clear indications are sent to external monitoring devices such as monitoring centers and handheld terminals used by maintenance personnel, either via wired or wireless means. This facilitates maintenance personnel's rapid understanding of the family-related defects in the industrial control system and timely intervention. It also provides accurate information for subsequently developing targeted maintenance strategies, procurement plans, design changes, or environmental modification schemes, thereby ensuring effective maintenance of the industrial control system and guaranteeing long-term operational reliability. Additionally, an audible and visual alarm system can be used; however, the specific design is based on existing technology and will not be elaborated upon here.
[0044] This targeted alarm directly transforms early warnings into actionable decision support, helping maintenance personnel move beyond the level of "where the fault is" and quickly focus on the root cause of "why the fault is," thereby providing a precise basis for subsequently developing the most targeted maintenance strategies, procurement plans, design changes, or environmental modification solutions, ultimately forming a closed-loop management system from intelligent diagnosis to precise implementation.
[0045] The above description is merely a preferred embodiment of the present invention and does not limit the patent scope of the present invention. Any equivalent structural or procedural transformations made based on the content of the present invention's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of the present invention.
Claims
1. A method for identifying family-related defects in a hydropower station control system, characterized in that, Includes the following steps: S1: Divide the hydropower station's industrial control system into multiple functional units and encode them digitally; S2: Construct a dual-identification feature vector set for each functional unit, and sort and normalize each feature vector in the dual-identification feature vector set; wherein, the feature vectors in the dual-identification feature vector set are arranged in the following order: defect severity S, defect occurrence frequency F, component batch consistency B, operating environment sensitivity E and design scheme reusability D, component supplier concentration V, component life loss rate L and component defect type consistency T. S3: The combined weighting method is used to calculate and determine the combined weight w of each feature vector in each functional unit; S4: Based on the feature vector data and weights of normal functional units without familial defects, calculate the cluster center vector in their healthy state as the identification benchmark vector. S5: Calculate the first relative Euclidean distance d1 between each feature vector in the target functional unit and its corresponding feature vector in the identification reference vector, and the second relative Euclidean distance d' between each feature vector in the target functional unit and its corresponding feature vector in each functional unit. ia Then, based on d1 and d' ia Calculate the familial defect difference d2 between any two adjacent functional units; The formula for calculating the first relative Euclidean distance d1 is: ； In the formula, k is the index of the eigenvector, and w k x is the combined weight of the k-th eigenvector in the target functional unit. k b is the normalized value of the k-th eigenvector of the target functional unit. 0k To identify the normalized value of the k-th eigenvector in the reference vector; Second relative Euclidean distance d' ia The formula for calculating the familial defect difference d2 is: ； ； In the formula, i is the sequence number of the target functional unit, a is the sequence number of each functional unit in the industrial control system, and i ≠ a; j is the sequence number of the feature vector in the functional unit; N is the number of functional units in the industrial control system; d' ia It is the second relative Euclidean distance between any two functional units; S6: Based on the historical feature vector data of functional units with familial defects and the historical feature vector data of normal functional units, construct the ROC working curve and make dynamic adjustments to obtain and correct the dynamic distance adjustment threshold y0 corresponding to different functional units. S7: Based on the first relative Euclidean distance d1 and the second relative Euclidean distance d' of each functional unit ia The familial defect difference degree d2 and the corresponding dynamic distance adjustment threshold y0 are used to determine the familial defect status of each functional unit. If d1 > y0 of the target functional unit, the functional unit is determined to have a single family defect; if no less than 3 functional units simultaneously satisfy d1 > y0 and d2 is less than the upper limit of the preset second Euclidean distance d2, the industrial control system is determined to have a cluster family defect. S8: For industrial control systems with clustered or family-like defects, trace the root cause of the defects based on the commonalities of the defects.
2. The method for identifying familial defects according to claim 1, characterized in that, In step S1, the functional units of the industrial control system include at least a power supply unit, a PLC controller unit, an execution drive unit, a key sensing element unit, and a communication interface unit.
3. The method for identifying familial defects according to claim 2, characterized in that, In step S2, the dual identification feature vector set is divided into a unit dimension vector subset and a component dimension vector subset; among them, the defect severity S, defect occurrence frequency F, component batch consistency B, operating environment sensitivity E, and design scheme reusability D belong to the unit dimension vector subset; the component supplier concentration V, component life loss rate L, and component defect type consistency T belong to the component dimension vector subset.
4. The method for identifying familial defects according to claim 3, characterized in that, In step S2, the feature vectors are constructed as follows: The severity of defects (S) is divided into five levels from low to high based on the severity of the consequences: no defects, false alarms, parameter drift, logic errors, and program crashes. The defect frequency F is calculated based on the number of defects occurring every 30 days, a. ; Component batch consistency B is calculated based on the proportion of components in the same batch to the total number of components; The sensitivity of the operating environment E is calibrated based on the increase in defect incidence when the operating environment temperature is >45℃, and the maximum value is taken when the defect incidence increases to two times or more. The reusability D of the design scheme is calculated based on the reuse ratio of the PLC control program. Supplier concentration V is calculated as the proportion of the number of core components within a functional unit that come from the same supplier to the total number of core components. The life loss rate L is calculated as the ratio of the remaining life of the component to its design life. Defect type consistency T is calculated based on the proportion of the number of defects of the same type in a functional unit to the total number of defects within a preset statistical period.
5. The method for identifying familial defects according to claim 4, characterized in that: The normalization method used in step S2 is specifically the maximum-minimum method.
6. The method for identifying familial defects according to claim 5, characterized in that, In step S2, within a single functional unit, each feature vector is normalized to form a dual-identification feature vector set, as shown below: ; Where X1 is the set of feature vectors of the first functional unit; x 11 The first dimension is the feature vector; Then, let X be the set containing the dual-identification feature vector sets of all functional units, and X is expressed as: 。 7. The method for identifying familial defects according to claim 1, characterized in that, The combined weighting method in step S3 is specifically as follows: For each feature vector in each functional unit, the subjective and objective weights are calculated using the analytic hierarchy process (AHP) and the entropy weighting method, respectively. Then, the final combined weight of the target feature vector is obtained through a combination operation, and its calculation formula is as follows: W k =Subjective weight × 0.6 + Objective weight × 0.
4.
8. The method for identifying familial defects according to claim 1, characterized in that, In step S6, the specific method for dynamically correcting the ROC curve is as follows: whenever 50 new sets of feature vector data with familial defects are added, the ROC curve is recalculated based on the new data and the historical feature vector data of the normal functional unit, and the corrected dynamic distance adjustment threshold y0 is obtained.
9. The method for identifying familial defects according to claim 1, characterized in that, In step S7, the preset upper limit of the second Euclidean distance threshold is 0.
3.
10. The method for identifying familial defects according to claim 3, characterized in that, In step S8, the method for tracing the root cause of the defect is specifically as follows: S81: Preset common defect weight threshold w 缺 In each functional unit exhibiting clustered or family-like defects, extract the combined weight value w. k Higher than w 缺 The eigenvectors are analyzed, and the root cause of familial defects is determined based on the combination of each eigenvector and its subset. S82: Based on the judgment result, send out different alarm messages.