A method and system for diagnosing and tracing oil failures based on multi-index monitoring

Abstract

This invention discloses a method and system for diagnosing and tracing oil failures based on multi-index monitoring. It constructs a tree-like structure for multi-index failure diagnosis; unifies the dimensions of oil monitoring data at the bottom-event index layer and divides it into state levels; uses fuzzy membership functions to fuzzify the dimension-unified index layer data, obtaining the membership probability of the oil monitoring data corresponding to each state level; establishes an expert system based on IF-THEN rules, formulating different inference rules for the AND and OR gates of the tree structure, and inputting the membership probabilities into the expert system based on IF-THEN rules to obtain the joint probability value of the oil's overall state belonging to each state level; assigns different utility intervals to different state levels of the oil based on the joint probability value, defuzzifies to obtain the comprehensive oil state value, and diagnoses the oil state based on a threshold; and traces the source of data diagnosed as failures. This invention improves the scientificity and accuracy of oil failure state determination.

Description

Technical Field

[0001] This invention belongs to the field of oil condition monitoring technology, specifically relating to a method and system for diagnosing and tracing oil failures based on multi-index monitoring. Background Technology

[0002] Lubricating oil circulates continuously in tribological systems, possessing strong failure early warning capabilities and conveying detailed information about system operation, thus reflecting the health status of machinery and its components. Oil monitoring is a technique for comprehensive analysis of the physicochemical properties of lubricating oil. Monitoring indicators requires specialized equipment for offline testing, which is both expensive and time-consuming. This results in limited monitoring data and long data acquisition intervals for lubricating oil indicators during machine operation. However, oil has a wide variety of detectable indicators, and multiple indicators can be analyzed and monitored for the same batch of oil. Therefore, comprehensive analysis using multiple oil indicators can compensate for the insufficient data from single-indicator monitoring. Furthermore, how to trace the source of failures in failed oil data to analyze specific equipment operating information is also a problem restricting the development of oil monitoring technology.

[0003] Although the amount of oil monitoring data during equipment operation is relatively small, the types of indicators monitored are diverse, mainly including physicochemical properties, abrasive particles, contaminants, and additives. Physicochemical properties mainly include lubricating oil oxidation, viscosity, acid value, and alkalinity, while lubricating oil contaminants mainly include moisture and particulate matter. The growth trends of various oil indicators are inconsistent and may even conflict. The introduction of expert knowledge can reduce information conflicts between indicators. By establishing a rule base, multiple pieces of expert knowledge can be fused to extract oil state characteristics, providing more scientific basis for multi-indicator oil monitoring.

[0004] In the process of oil monitoring, it is necessary not only to perform relevant performance analysis and failure diagnosis on the oil monitoring indicators, but also, since the ultimate goal of analyzing oil performance indicators is to monitor the operating status of equipment, it is necessary to perform reverse analysis on the oil failure data of equipment operation to determine the oil indicators that caused the failure, so as to facilitate rapid diagnosis and maintenance. This requires failure tracing in oil monitoring data.

[0005] Therefore, how to perform oil condition analysis with limited oil data, i.e., multi-index analysis of oil, and how to analyze and trace equipment failure data are problems that oil monitoring needs to solve. Summary of the Invention

[0006] The technical problem to be solved by the present invention is to provide a method and system for diagnosing and tracing oil failure based on multi-index monitoring, which addresses the shortcomings of the prior art and improves the accuracy of oil condition diagnosis and enables the tracing of failure data.

[0007] The present invention adopts the following technical solution:

[0008] A method for diagnosing and tracing oil failures based on multi-index monitoring includes the following steps:

[0009] S1. Construct a multi-index failure diagnosis tree structure, with each layer of the multi-index failure diagnosis tree structure connected by AND gates or OR gates;

[0010] S2. The oil monitoring data of the bottom event index layer in the multi-index failure diagnosis tree structure constructed in step S1 is unified in terms of dimensions and divided into state levels. The index layer data after dimension unification is fuzzified by using fuzzy membership function. The index is evaluated to different state levels to obtain the membership probability of oil monitoring data for each state level.

[0011] S3. Establish an expert system based on IF-THEN rules, formulate different inference rules for the AND and OR gates of the tree structure in step S1, input the membership probability obtained in step S2 into the expert system based on IF-THEN rules, perform multi-index fusion of oil in the faulty tree structure, and obtain the joint probability value of the comprehensive state of oil belonging to each state level.

[0012] S4. Assign different utility ranges to different state levels of oil based on the joint probability value obtained in step S3, defuzzify to obtain the comprehensive state value of oil, set the threshold ε of the comprehensive state value based on actual data and relevant oil monitoring standards, and diagnose the state of oil based on the threshold ε.

[0013] S5. Trace the source of the data diagnosed as invalid in step S4.

[0014] Specifically, in step S1, the multi-index failure diagnosis tree structure includes four layers. The bottom events of the fourth layer serve as the input to the multi-index failure diagnosis tree structure and are used to map oil indicators. The third and second layers are intermediate events and are used to map oil properties. The first layer is the top event and is used to map the state of the oil. The first and second layers are connected by an OR gate, and the second and third layers, as well as the third and fourth layers, are connected by an AND gate or an OR gate.

[0015] Specifically, in step S2, the dimensions of the indicator layer data are unified to obtain dimensionless normalized data. The status levels are divided into N levels, and fuzzy membership functions are used to calculate the membership degree of the monitoring data for each status level.

[0016] Furthermore, normalized data for:

[0017]

[0018] Where i,j=1,2,3……,a ijN Indicator a ij The threshold for failure, a ij0 Indicator a ij Initial value, The data represents the normalized oil performance indicators. I1 is the set of benefit-type indicators, and I2 is the set of cost-type indicators.

[0019] Specifically, in step S3, different inference rules are formulated for the AND and OR gates of the tree structure in step S1. Starting from the fourth layer monitoring index of the tree structure, the attribute part of the third layer is obtained by applying the AND gate inference rule library. In the third layer, different inference rule libraries are applied to the three branches of the tree structure to obtain the oil attribute part of the second layer. Through the second layer OR gate inference rule library, the joint probability of the first layer oil comprehensive state belonging to each state level is obtained.

[0020] Furthermore, the k-th inference rule is as follows:

[0021] IF:

[0022] THEN:{(H1,β1),…,(H c ,β c ),…,(H N ,β N )

[0023] In this context, the IF part represents the antecedent of the rule, and the THEN part represents the consequent of the rule. H represents the i-th oil data in the antecedent of the k-th rule, where i,k = 1,2,3,... c For the state level, r represents the number of data points; β c For the reliability in the consequent of the rule, c represents the c-th level, N is the total number of state levels, and H is the reliability. N For the Nth state level, β N This represents the probability of belonging to the Nth state level.

[0024] Furthermore, the oil multi-indicator fusion is performed along the fault tree structure, specifically as follows:

[0025] Based on the inference rules of AND and OR gates in a tree structure, four rule bases are established, including two-attribute and three-attribute inference rules for AND gates, and two-attribute and three-attribute inference rules for OR gates. After the rule bases are established, at the bottom level of the tree structure, for index a... 41 ,a 42 Fuzzy reasoning is performed using the two-input Rule4 rule base to obtain attribute A4, for index a. 51 ,a 52Fuzzy reasoning is performed using the two-input Rule3 rule base to obtain attribute A5. In the third layer, for index a... 11 ,a 12 Given attribute A4, fuzzy reasoning is performed using the three-input Rule1 rule base to obtain attribute A1, for index a. 21 ,a 22 Fuzzy reasoning is performed using the two-input Rule3 rule base to obtain attribute A2, for index a. 31 ,a 32 Given attribute A5, we apply the three-input Rule2 rule base for fuzzy reasoning to obtain attribute A2. In the second layer, for attributes A1, A2, and A3, we apply the three-input Rule1 rule base for fuzzy reasoning to obtain the joint probability of the first layer oil comprehensive state belonging to each state level.

[0026] Specifically, in step S4, based on the joint probability values ​​of each state level of the oil obtained in step S3, the oil state level is assigned a corresponding utility interval, and defuzzification is performed to obtain the comprehensive state evaluation result HI of the oil. When HI is less than the comprehensive state judgment threshold ε, the oil is diagnosed as healthy. When HI is greater than or equal to the comprehensive state judgment threshold ε, the oil is diagnosed as failed.

[0027] Specifically, in step S5, tracing the source of data whose diagnostic result is invalid involves the following steps:

[0028] When the index value of the oil is greater than or equal to the index threshold, the corresponding oil index is defined as the oil index that caused the failure, and the source tracing ends.

[0029] For efficiency-related indicators, when the values ​​of various indicators of the oil are equal to or lower than the indicator threshold, the corresponding oil indicator is defined as the oil indicator that caused the failure, and the source tracing ends.

[0030] When the index value of the oil is less than the index threshold, the attribute layer data of the oil is defuzzified starting from the second layer. The defuzzified values ​​are compared. For the attribute with the largest value, the failure source is traced along the corresponding attribute in the third layer branch. If the corresponding attribute in the third layer branch is all index data, the index data after unification of dimensions under the corresponding branch is compared, and the maximum value is set as the oil index that caused the failure.

[0031] When the corresponding attribute has both indicator data and attribute data in the third-level branch, the attribute data is defuzzified and compared with the unified-dimensional indicator data in the corresponding branch. If the indicator data is the largest, the corresponding indicator is defined as the oil indicator that caused the failure. If the defuzzified attribute data is the largest, the source is traced back to the fourth-level branch.

[0032] For the index data branch with attributes in the fourth layer, compare the index data with the unified dimensions under the corresponding branch, and set the maximum value as the oil index that caused the failure, thus completing the failure tracing.

[0033] Secondly, embodiments of the present invention provide an oil failure diagnosis and tracing system based on multi-index monitoring, comprising:

[0034] The structural module constructs a tree structure for multi-index failure diagnosis, and the layers of the multi-index failure diagnosis tree structure are connected by AND gates or OR gates.

[0035] The module is divided into sections. The oil monitoring data of the bottom event index layer in the multi-index failure diagnosis tree structure constructed by the structure module is unified in terms of dimensions and divided into state levels. Fuzzy membership functions are used to fuzzify the index layer data after the dimension unification. The membership evaluation of the index to different state levels is carried out to obtain the membership probability of the oil monitoring data for each state level.

[0036] The expert module establishes an expert system based on the IF-THEN rule. Different inference rules are formulated for the AND and OR gates of the tree structure of the structural module. The membership probability obtained by dividing the module is input into the expert system based on the IF-THEN rule. The faulty tree structure is fused with multiple oil indicators to obtain the joint probability value of the comprehensive state of the oil belonging to each state level.

[0037] The diagnostic module assigns different utility ranges to different state levels of the oil based on the joint probability values ​​obtained from the expert module, defuzzifies to obtain the comprehensive state value of the oil, sets a threshold ε for the comprehensive state value based on actual data and relevant oil monitoring standards, and diagnoses the state of the oil based on the threshold ε.

[0038] The traceability module traces the source of data that has been diagnosed as invalid by the diagnostic module.

[0039] Compared with the prior art, the present invention has at least the following beneficial effects:

[0040] This invention presents a multi-index monitoring-based method for diagnosing and tracing the source of oil failure. It establishes a multi-index failure diagnosis tree structure incorporating expert knowledge and evidence-based reasoning. The method categorizes lubricating oil based on its physicochemical state, additive state, contaminant state, and wear state. Each attribute includes one or more indicators with the same characteristics, and different state levels are defined. A corresponding AND and OR gate reasoning rule base is established based on expert knowledge. Through evidence-based reasoning methods, the reliability of the oil's comprehensive state belonging to each state level is obtained. The state level is assigned a value and defuzzified to obtain the comprehensive oil state value. A comprehensive oil state threshold is then determined to identify whether the oil has failed, and the failure is traced back to its source. By monitoring multiple oil indicators, the uncertainty in diagnostic results caused by limited monitoring data and long data intervals is reduced, resulting in more accurate and reliable oil state assessments. Furthermore, for failed oil data, starting from the first layer of the comprehensive oil state layer, the method traces the source of failure layer by layer along the multi-index failure diagnosis tree structure to identify the oil indicator causing the failure, facilitating rapid diagnosis and repair.

[0041] Furthermore, based on expert experience and the mechanism of oil condition degradation, numerous monitoring indicators reflecting oil condition are divided into three attribute sets. Each attribute set contains several oil indicators and smaller attribute sets, forming a tree structure based on the correlation between the oil indicators. By integrating multiple oil monitoring indicators with the failure tree structure, on the one hand, the indicators are further subdivided, and more flexible expert rules are formulated for integration; on the other hand, during failure tracing, the path of oil indicators affecting the relevant properties of the oil and ultimately leading to oil condition failure can be more intuitively seen, facilitating subsequent failure tracing.

[0042] Furthermore, the dimensions of the multi-indicator data from oil monitoring are standardized to facilitate subsequent multi-indicator fusion monitoring of the oil. Fuzzy membership functions are used to obtain the membership probability of each oil data point belonging to various state levels, rather than belonging to a single state level, thus increasing the accuracy of indicator fusion.

[0043] Furthermore, a reasoning rule base is established based on expert experience. Flexible and reliable expert rules are set up for different oil indicators and different fusion situations. Through expert rules and evidence reasoning, the failure diagnosis tree structure is fused layer by layer. The feature dimension of the indicator data is reduced layer by layer through fusion. In the fusion process, expert rules and the mechanism of oil degradation are embedded, which improves the accuracy and interpretability of oil failure diagnosis.

[0044] Furthermore, the oil state levels are assigned utility intervals for defuzzification, and the joint probability of the comprehensive oil state belonging to each state level obtained in step S3 is transformed into a precise value of the comprehensive oil state evaluation result HI located in the interval [0,1]. The result of multi-index fusion of oil is more intuitive, making it easier to set thresholds for comprehensive state and determine failure.

[0045] Furthermore, by conducting source analysis on failure data, we can obtain specific indicator information that affects and causes oil failure, and then indirectly understand the operating status of the equipment at this time through the indicators, providing a more scientific basis for equipment maintenance.

[0046] It is understood that the beneficial effects of the second and third aspects mentioned above can be found in the relevant descriptions in the first aspect mentioned above, and will not be repeated here.

[0047] In summary, this invention improves the scientific rigor and accuracy of oil failure condition determination, and provides a more scientific basis for oil condition analysis and equipment maintenance by conducting source tracing analysis of failure data.

[0048] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0049] Figure 1 This is a tree structure diagram of the multi-index failure diagnosis of the present invention;

[0050] Figure 2 This is a schematic diagram of the failure diagnosis and tracing process of the present invention;

[0051] Figure 3 Verification diagram for the multi-indicator fusion model

[0052] Figure 4 This is a diagram used to verify an example of failure tracing. Detailed Implementation

[0053] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0054] In the description of this invention, it should be understood that the terms "comprising" and "including" indicate the presence of the described features, integrals, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or collections thereof.

[0055] It should also be understood that the terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the invention. As used in this specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.

[0056] It should also be further understood that the term "and / or" as used in this specification and the appended claims refers to any combination and all possible combinations of one or more of the associated listed items, and includes such combinations. For example, A and / or B can represent three cases: A alone, A and B simultaneously, and B alone. Additionally, the character " / " in this document generally indicates that the preceding and following objects have an "or" relationship.

[0057] It should be understood that although terms such as first, second, third, etc., may be used in the embodiments of the present invention to describe the preset range, these preset ranges should not be limited to these terms. These terms are only used to distinguish the preset ranges from one another. For example, without departing from the scope of the embodiments of the present invention, the first preset range may also be referred to as the second preset range, and similarly, the second preset range may also be referred to as the first preset range.

[0058] Depending on the context, the word "if" as used here can be interpreted as "when," "when," "in response to determination," or "in response to detection." Similarly, depending on the context, the phrase "if determination" or "if detection (of the stated condition or event)" can be interpreted as "when determination," "in response to determination," "when detection (of the stated condition or event)," or "in response to detection (of the stated condition or event)."

[0059] The accompanying drawings illustrate various structural schematic diagrams according to embodiments disclosed in this invention. These drawings are not to scale, and some details have been enlarged for clarity, and some details may have been omitted. The shapes of the various regions and layers shown in the drawings, as well as their relative sizes and positional relationships, are merely exemplary and may deviate from reality due to manufacturing tolerances or technical limitations. Furthermore, those skilled in the art can design regions / layers with different shapes, sizes, and relative positions as needed.

[0060] This invention provides a multi-index monitoring-based method for diagnosing and tracing the source of oil failures. It establishes a multi-index failure diagnosis tree structure incorporating expert knowledge and evidence-based reasoning. The method categorizes lubricating oil based on its physicochemical state, additive state, contaminant state, and wear state. Each attribute includes one or more indicators with the same characteristics, and different state levels are defined. A corresponding AND and OR gate reasoning rule base is established based on expert knowledge. Fuzzy reasoning and evidence synthesis methods are used to obtain the confidence level of the oil's comprehensive state belonging to each state level. The state level is assigned a value and defuzzified to obtain the comprehensive oil state value. A comprehensive oil state threshold is then determined to identify whether equipment failure exists and to trace the source of the failure. By monitoring multiple oil indicators, the uncertainty of diagnostic results caused by insufficient lubricating oil indicator monitoring data and long data acquisition intervals is reduced, resulting in more accurate and reliable oil state assessment results. Furthermore, for failed oil data, starting from the first layer of the comprehensive oil state layer, the method traces the source of failure layer by layer along the multi-index failure diagnosis tree structure to identify the oil indicator causing the failure, facilitating rapid diagnosis and repair.

[0061] This invention discloses a method for diagnosing and tracing oil failures based on multi-index monitoring, comprising the following steps:

[0062] S1. Construct a multi-index failure diagnosis tree structure. Circles represent the bottom events of the tree structure, which map the indicators of the oil and are the inputs of the tree structure. The boxes in the second and third layers represent the intermediate events of the tree structure, which map the properties of the oil. The boxes in the first layer represent the top events of the tree structure, which map the state of the oil and are used for subsequent failure diagnosis of the oil state.

[0063] And represents an AND gate in a tree structure, and Or represents an OR gate in a tree structure, each representing a different inference rule base. The layers of the tree structure are connected by AND and OR gates, indicating that the properties and states of the oil are obtained through different inference rules.

[0064] Please see Figure 1 The bottom event circles in the tree structure represent oil monitoring data acquired through various monitoring methods, serving as the input for the tree structure. The middle event boxes represent attributes categorized according to the lubricating oil's physicochemical state, additive state, contaminant state, and wear state. Each attribute contains one or more indicators with the same characteristics. For example, moisture and zinc content constitute the additive state attribute, while iron and copper content constitute the oil wear state attribute, and so on.

[0065] Starting from the fourth layer, the circular portion is used to obtain the box portion of the third layer through the corresponding inference rules of the AND and OR gates. The circular and box portions of the third layer are then used to obtain the box portion of the second layer through the corresponding AND and OR gate inference rules. The box portion of the second layer is then used to obtain the oil state of the top event of the first layer through the corresponding OR gate inference rules. The oil state of the top event in the tree structure is the final output, used for subsequent failure diagnosis of the oil state. The comprehensive state value HI of the oil ranges from [0,1], where 0 indicates that the oil is in the best state and 1 indicates that the oil is in a completely failed state.

[0066] S2. First, the oil monitoring data of the bottom event is standardized in terms of dimensions, that is, the data is normalized and divided into state levels. Then, the index data is fuzzified by using fuzzy membership functions, and the index is evaluated to belong to different state levels.

[0067] Oil monitoring indicators obtained through other monitoring methods are divided into benefit-oriented indicators and cost-oriented indicators. Benefit-oriented indicators are those with higher values, such as zinc content in additives and total base number (TBN). Cost-oriented indicators are those with lower values, such as viscosity change rate, acid number (TAN) change, and contaminant content.

[0068] For the obtained multi-index monitoring data, since the dimensions and magnitudes of the time series of different oil indicators are inconsistent, equation (1) can be used to unify the dimensions of the monitoring index data to obtain dimensionless normalized data.

[0069]

[0070] Where i,j=1,2,3……,a ijN Indicator a ij The failure threshold can be set with reference to the failure values ​​specified in the oil change standard; a ij0 Indicator a ij The initial value can be set with reference to the new oil index. The data represents the normalized oil performance indicators. I1 is the set of benefit-type indicators, and I2 is the set of cost-type indicators.

[0071] After normalization, the state levels are divided according to the properties of the oil, from best to worst, into {H1, H2, ..., H...}. c …,H N There are N levels.

[0072] To assess the state level corresponding to the indicators, fuzzy membership functions are used to calculate the membership degree of the monitoring indicator data for each state level, thus obtaining the membership probability of the monitoring data for each state level.

[0073] In this invention, the fuzzy membership function is selected as a Gaussian function.

[0074] S3. Establish an expert system based on IF-THEN rules, formulate different inference rules for AND and OR gates in the tree structure, and start from the fourth level of the tree structure, inputting the monitoring index a obtained in S2. 41 and a 42 a 51 and a 52 For the membership probabilities of each state level, the AND and OR gate inference rule bases are applied to obtain the membership probabilities of A4 and A5 in the third layer for each state level. Different inference rule bases are applied to the three branches of the tree structure in the third layer to obtain the membership probabilities of A1, A2 and A3 in the second layer for each state level. Through the OR gate inference rule base in the second layer, the joint probability of the oil liquid comprehensive state belonging to each state level is obtained.

[0075] The expert system based on the IF-THEN rule formulates different inference rules for AND and OR gates in a tree structure, where the k-th inference rule is shown below:

[0076] IF:

[0077] THEN:{(H1,β1),…,(H c ,β c ),…,(H N ,β N )

[0078] In the rule, the IF part is the antecedent of the rule, and the THEN part is the consequent of the rule. This represents the i-th oil data in the antecedent of the k-th rule, where i,k = 1, 2, 3, ..., referring to the oil index a. ij It also refers to attribute data A. i It belongs to state level H c The probability can be calculated using the Gaussian membership function, where r represents the number of data points; β c The confidence level in the consequent of a rule indicates that the state of the oil obtained by reasoning under this rule belongs to state level H. c The probability is set based on expert experience and knowledge of this rule, where c = 1, 2, 3... represents the c-th level, N is the total number of state levels, and H... N For the Nth state level, β N This represents the probability of belonging to the Nth state level.

[0079] When using the inference rule base for inference fusion, the input is the probability of the oil index or the attribute to belong to each state level. During fusion, the output of each rule in the rule base, i.e. the consequent of the rule, needs to be multiplied by the corresponding activation weight and weighted to obtain the final result of evidence inference. Then, multiple fusions are performed along the failure diagnosis tree through multiple rule bases to finally obtain the joint probability of the comprehensive state of the oil belonging to each state level.

[0080] Based on the different reasoning rules of AND and OR gates in a tree structure, four rule bases were established, mainly including two-input and three-input reasoning rules for AND gates and two-input and three-input reasoning rules for OR gates. The location and usage of the rule bases are as follows: Figure 1 As shown in Rule 1, Rule 2, Rule 3, and Rule 4.

[0081] Please see Figure 3 and Figure 4 After the rule base is established, at the bottom level of the tree structure, for the monitoring indicator a obtained in step S2 41 ,a 42 The membership probability of attribute A4 to each state level is obtained by applying evidence reasoning to the two-input Rule4 rule base. This yields the membership probability of attribute A4 to each state level. For the monitoring indicator a obtained in S2... 51 ,a 52 The membership probability of attribute A5 to each state level is obtained by applying evidence reasoning to the two-input Rule3 rule base. In the third layer, for the monitoring index a obtained in step S2... 11 ,a 12 Given the membership probability of attribute A4 to each state level, we apply the three-input Rule1 rule base for evidence reasoning to obtain the membership probability of attribute A1 to each state level. This is then used for the monitoring index a obtained in step S2. 21 ,a 22 The membership probability of attribute A2 to each state level is obtained by applying evidence reasoning to the two-input Rule3 rule base. This yields the membership probability of attribute A2 to each state level. For the monitoring index a obtained in step S2... 31 ,a 32 Given the membership probability of attribute A5 to each state level, we apply the three-input Rule2 rule base for evidence reasoning to obtain the membership probability of attribute A2 to each state level. In the second layer, for the membership probabilities of attributes A1, A2, and A3 to each state level, we apply the three-input Rule1 rule base for evidence reasoning to obtain the joint probability of the first layer oil liquid comprehensive state to each state level.

[0082] S4. Based on experience, different utility ranges are assigned to different state levels of the oil. The comprehensive state value of the oil is obtained by defuzzification. The threshold ε of the comprehensive state value is set according to actual data and relevant oil monitoring standards. The results are compared and the multi-index failure diagnosis results of the oil are output.

[0083] Based on the joint probability value of the oil's overall state belonging to each state level obtained in step S3, the corresponding utility interval is assigned to the oil's state level, and the fuzziness is de-fuzzified using equation (2) to obtain the overall state evaluation result of the oil.

[0084]

[0085] Where HI(t) is the comprehensive oil condition assessment result at time t, β c (t) represents the overall oil condition level at time t, which is H. c The reliability of H, i.e., membership in H c The membership probability value of the level, μ(H) c The assessment status level is H. c The utility intervals, whose parameter values ​​need to be obtained through training with monitoring data, c = 1, 2, 3...

[0086] Through simulation calculations and analysis of actual monitoring data, combined with relevant standards for oil monitoring, a suitable threshold was finally determined. The threshold for judging the overall state during multi-index oil monitoring was set as ε. When HI is less than ε, the oil is diagnosed as healthy; when HI is greater than or equal to ε, the oil is diagnosed as failing.

[0087] S5. Failure tracing: In step S4, after performing failure diagnosis on the multi-index data of the oil, data with a diagnosis result of "healthy" are not traced. Data with a diagnosis result of "failure" are traced.

[0088] Please see Figure 2 First, determine whether the values ​​of each oil indicator exceed the indicator threshold. The indicator threshold refers to the threshold at which the indicator fails. It is formulated according to the industrial oil change standard. For cost-related indicators, if they are greater than or equal to the indicator threshold, the indicator is directly identified as the oil indicator that caused the failure, and the traceability ends. For benefit-related indicators, if they are equal to or lower than the indicator threshold, the indicator is directly identified as the oil indicator that caused the failure, and the traceability ends.

[0089] If the values ​​of each index of the oil are less than the index threshold, the attribute layer data of the oil is defuzzified starting from the second layer. The defuzzification process is similar to step S4. The state level to which the oil attribute belongs is assigned a corresponding utility range, and attribute defuzzification is performed using equation (2). The defuzzified values ​​are compared. For the attribute with the largest value, failure tracing continues along the branch of the third layer. If the branch of the attribute in the third layer is all index data, the normalized unified dimension monitoring index data under the branch is compared, and the maximum value is set as the oil index that caused the failure. If the branch of the attribute in the third layer has both index data and attribute data, the attribute data is defuzzified and compared with the unified dimension index data under the branch. If the index data is the largest, the index is set as the oil index that caused the failure. If the attribute data after defuzzification is the largest, the tracing continues along the branch of the fourth layer. For the index data branch of the attribute in the fourth layer, the unified dimension index data under the branch is compared, and the maximum value is set as the oil index that caused the failure, thus completing the failure tracing.

[0090] The source tracing example is as follows: If all the monitored indicator data are less than the threshold, then starting from the second layer, use equation (2) to defuzzify attributes A1, A2, and A3 respectively, and then compare attributes A1, A2, and A3. For the attribute with the largest value, continue to trace the source. If the value of attribute A1 is the largest, then in the third layer, use equation (2) to defuzzify attribute A4, and then compare the normalized indicator a. 11 ,a 12 And attribute A4, if index a 11 or a 12 If the value is the largest, the corresponding indicator will be identified as the source of failure, and the source will be traced.

[0091] If attribute A4 is the largest, then continue tracing back to the source and compare the normalized index a. 41 ,a 42 The indicator with the largest value is identified as the source of failure, thus completing the source tracing. The source tracing process for other attributes or indicators is similar to that for A1, and will not be elaborated upon.

[0092] In another embodiment of the present invention, an oil failure diagnosis and tracing system based on multi-index monitoring is provided. This system can be used to implement the above-mentioned oil failure diagnosis and tracing method based on multi-index monitoring. Specifically, the oil failure diagnosis and tracing system based on multi-index monitoring includes a structural module, a partitioning module, an expert module, a diagnosis module, and a tracing module.

[0093] Among them, the structural module constructs a multi-index failure diagnosis tree structure, and the layers of the multi-index failure diagnosis tree structure are connected by AND gates or OR gates.

[0094] The module is divided into sections. The oil monitoring data of the bottom event index layer in the multi-index failure diagnosis tree structure constructed by the structure module is unified in terms of dimensions and divided into state levels. Fuzzy membership functions are used to fuzzify the index layer data after the dimension unification. The membership evaluation of the index to different state levels is carried out to obtain the membership probability of the oil monitoring data for each state level.

[0095] The expert module establishes an expert system based on the IF-THEN rule. Different inference rules are formulated for the AND and OR gates of the tree structure of the structural module. The membership probability obtained by dividing the module is input into the expert system based on the IF-THEN rule. The faulty tree structure is fused with multiple oil indicators to obtain the joint probability value of the comprehensive state of the oil belonging to each state level.

[0096] The diagnostic module assigns different utility ranges to different state levels of the oil based on the joint probability values ​​obtained from the expert module, defuzzifies to obtain the comprehensive state value of the oil, sets a threshold ε for the comprehensive state value based on actual data and relevant oil monitoring standards, and diagnoses the state of the oil based on the threshold ε.

[0097] The traceability module traces the source of data that has been diagnosed as invalid by the diagnostic module.

[0098] In summary, this invention provides a multi-indicator monitoring-based method and system for diagnosing and tracing oil failures. Addressing the issue of limited monitoring data and long data acquisition intervals leading to significant uncertainty in oil condition monitoring, this invention utilizes multiple oil monitoring indicators to establish a multi-indicator failure diagnosis tree structure incorporating expert knowledge and evidence-based reasoning. This results in more accurate and reliable oil condition assessments. Furthermore, for data on failed oil, starting from the first layer (comprehensive oil condition layer), the system traces the failure layer by layer along the multi-indicator failure diagnosis tree structure to identify the specific oil indicator causing the failure, facilitating rapid diagnosis and repair.

[0099] The above content is only for illustrating the technical concept of the present invention and should not be construed as limiting the scope of protection of the present invention. Any modifications made to the technical solution based on the technical concept proposed in this invention shall fall within the scope of protection of the claims of this invention.

Claims

1. A method for diagnosing and tracing the source of oil failure based on multi-index monitoring, characterized in that, Includes the following steps: S1. Construct a multi-index failure diagnosis tree structure, with each layer of the multi-index failure diagnosis tree structure connected by AND gates or OR gates; S2. The oil monitoring data of the bottom event index layer in the multi-index failure diagnosis tree structure constructed in step S1 is dimensionally unified and divided into state levels. Fuzzy membership functions are used to fuzzify the dimensionally unified index layer data, and the membership of the indexes to different state levels is evaluated to obtain the membership probability of the oil monitoring data for each state level. Dimensionally unified index layer data is then obtained to obtain dimensionless normalized data. ; The status levels are divided into The index data is calculated using fuzzy membership functions at each level to obtain the membership degree of the monitoring data for each state level. Normalized data for: in, , Indicators The threshold for failure Indicators Initial value, These are the normalized oil performance data. It is a set of benefit-oriented indicators. It is a set of cost-related indicators; S3. Establish an expert system based on IF-THEN rules. Formulate different inference rules for the AND and OR gates of the tree structure in step S1. Input the membership probability obtained in step S2 into the expert system based on IF-THEN rules. Perform multi-index fusion of oil in the faulty tree structure to obtain the joint probability value of the oil comprehensive state belonging to each state level. Formulate different inference rules for the AND and OR gates of the tree structure. Starting from the fourth layer monitoring index of the tree structure, apply the AND gate inference rule library to obtain the attribute part of the third layer. Apply different inference rule libraries to the three branches of the tree structure in the third layer to obtain the second layer oil attribute part. Through the second layer OR gate inference rule library, obtain the joint probability of the first layer oil comprehensive state belonging to each state level. No. k The following are the rules of inference: IF: is and … is and … and is THEN: In this context, the IF part represents the antecedent of the rule, and the THEN part represents the consequent of the rule. Indicates the first k The first clause of the preceding rule i Individual oil data, , For status level, r Represents the number of data points; For the reliability in the consequent of the rule, Indicates the first Each level N The total number of state levels. For the first N Each status level To belong to the N The probability of each state level; The oil performance multi-indicator fusion is performed along the fault tree structure, specifically as follows: Based on the inference rules of AND and OR gates in a tree structure, four rule bases are established, including two-attribute and three-attribute inference rules for AND gates, and two-attribute and three-attribute inference rules for OR gates. After the rule bases are established, at the bottom level of the tree structure, for indicators... Fuzzy reasoning is performed using a two-input Rule4 rule base to obtain attributes. For indicators Fuzzy reasoning is performed using a two-input Rule3 rule base to obtain attributes. In the third layer, for indicators and attributes Fuzzy reasoning is performed using a three-input Rule1 rule base to obtain attributes. For indicators Fuzzy reasoning is performed using a two-input Rule3 rule base to obtain attributes. For indicators and attributes Fuzzy reasoning is performed using a three-input Rule2 rule base to obtain attributes. In the second layer, for attributes , , Fuzzy reasoning is performed using the three-input Rule1 rule base to obtain the joint probability of the first layer of oil comprehensive state belonging to each state level; S4. Based on the joint probability values ​​obtained in step S3, assign different utility intervals to different state levels of the oil, defuzzify to obtain the comprehensive state value of the oil, set a threshold ε for the comprehensive state value based on actual data and relevant oil monitoring standards, diagnose the oil state based on the threshold ε, and assign corresponding utility intervals to the oil state levels based on the joint probability values ​​of each state level obtained in step S3, defuzzify to obtain the comprehensive state assessment result of the oil. ,when When the oil level is less than the comprehensive condition judgment threshold ε, the oil is diagnosed as being in a healthy state. When the overall condition judgment threshold ε is greater than or equal to the oil condition judgment threshold, the oil is diagnosed as being in a failure state. S5. Trace the source of the data diagnosed as invalid in step S4, specifically as follows: For cost-related indicators, when the indicator value of the oil is greater than or equal to the indicator threshold, the corresponding oil indicator is defined as the oil indicator that caused the failure, and the traceability ends. When the index value of the oil is less than the index threshold, the attribute layer data of the oil is defuzzified starting from the second layer. The defuzzified values ​​are compared. For the attribute with the largest value, the failure source is traced along the corresponding attribute in the third layer branch. If the corresponding attribute in the third layer branch is all index data, the index data after unification of dimensions under the corresponding branch is compared, and the maximum value is set as the oil index that caused the failure. When the corresponding attribute has both indicator data and attribute data in the third-level branch, the attribute data is defuzzified and compared with the unified-dimensional indicator data in the corresponding branch. If the indicator data is the largest, the corresponding indicator is defined as the oil indicator that caused the failure. If the defuzzified attribute data is the largest, the source is traced back to the fourth-level branch. For the index data branch with attributes in the fourth layer, compare the index data with the unified dimensions under the corresponding branch, and set the maximum value as the oil index that caused the failure, thus completing the failure source tracing. For efficiency-related indicators, when the values ​​of various indicators of the oil are equal to or lower than the indicator threshold, the corresponding oil indicator is defined as the oil indicator that caused the failure, and the traceability ends.

2. The oil failure diagnosis and tracing method based on multi-index monitoring according to claim 1, characterized in that, In step S1, the multi-index failure diagnosis tree structure includes four layers. The bottom events of the fourth layer serve as the input to the multi-index failure diagnosis tree structure and are used to map oil indicators. The third and second layers are intermediate events and are used to map oil properties. The first layer is the top event and is used to map the state of the oil. The first and second layers are connected by an OR gate, and the second and third layers, as well as the third and fourth layers, are connected by an AND gate or an OR gate.

3. A multi-index monitoring-based oil failure diagnosis and traceability system, characterized in that, include: The structural module constructs a tree structure for multi-index failure diagnosis, and the layers of the multi-index failure diagnosis tree structure are connected by AND gates or OR gates. The module is divided into sections. The oil monitoring data in the bottom event index layer of the multi-index failure diagnosis tree structure constructed by the structural module is dimensionally unified and divided into state levels. Fuzzy membership functions are used to fuzzify the dimensionally unified index layer data. Membership evaluation of the indicators to different state levels is performed to obtain the membership probability of the oil monitoring data for each state level. Dimensionally unified index layer data is then obtained to obtain dimensionless normalized data. ; Divide the status levels into The index data is calculated using fuzzy membership functions at each level to obtain the membership degree of the monitoring data for each state level. Normalized data for: in, , Indicators The threshold for failure Indicators Initial value, These are normalized oil performance data. It is a set of benefit-oriented indicators. It is a set of cost-related indicators; The expert module establishes an expert system based on the IF-THEN rule. Different inference rules are formulated for the AND and OR gates of the tree structure of the structural module. The membership probabilities obtained from dividing the modules are input into the expert system based on the IF-THEN rule. The faulty tree structure is fused with multiple oil indicators to obtain the joint probability value of the oil comprehensive state belonging to each state level. Different inference rules are formulated for the AND and OR gates of the tree structure. Starting from the fourth layer monitoring indicators of the tree structure, the AND gate inference rule library is applied to obtain the attribute part of the third layer. Different inference rule libraries are applied to the three branches of the tree structure in the third layer to obtain the oil attribute part of the second layer. Through the second layer OR gate inference rule library, the joint probability of the first layer oil comprehensive state belonging to each state level is obtained. No. k The following are the rules of inference: IF: is and … is and … and is THEN: In this context, the IF part represents the antecedent of the rule, and the THEN part represents the consequent of the rule. Indicates the first k The first clause of the preceding rule i Individual oil data, , For status level, r Represents the number of data points; For the reliability in the consequent of the rule, Indicates the first Each level N The total number of state levels. For the first N Each status level To belong to the N The probability of each state level; The oil performance multi-indicator fusion is performed along the fault tree structure, specifically as follows: Based on the inference rules of AND and OR gates in a tree structure, four rule bases are established, including two-attribute and three-attribute inference rules for AND gates, and two-attribute and three-attribute inference rules for OR gates. After the rule bases are established, at the bottom level of the tree structure, for indicators... Fuzzy reasoning is performed using a two-input Rule4 rule base to obtain attributes. For indicators Fuzzy reasoning is performed using a two-input Rule3 rule base to obtain attributes. In the third layer, for indicators and attributes Fuzzy reasoning is performed using a three-input Rule1 rule base to obtain attributes. For indicators Fuzzy reasoning is performed using a two-input Rule3 rule base to obtain attributes. For indicators and attributes Fuzzy reasoning is performed using a three-input Rule2 rule base to obtain attributes. In the second layer, for attributes , , Fuzzy reasoning is performed using the three-input Rule1 rule base to obtain the joint probability of the first layer of oil comprehensive state belonging to each state level; The diagnostic module assigns different utility intervals to different state levels of the oil based on the joint probability values ​​obtained from the expert module. It then defuzzifies the data to obtain a comprehensive oil state value. A threshold ε for this comprehensive state value is set based on actual data and relevant oil monitoring standards. The oil state is then diagnosed based on this threshold ε. Specifically, based on the obtained joint probability values ​​for each state level, a corresponding utility interval is assigned to each state level, and defuzzification is performed to obtain the comprehensive oil state assessment result. ,when When the oil level is less than the comprehensive condition judgment threshold ε, the oil is diagnosed as being in a healthy state. When the overall condition judgment threshold ε is greater than or equal to the oil condition judgment threshold, the oil is diagnosed as being in a failure state. The traceability module traces the source of data diagnosed as invalid by the diagnostic module. Specifically: For cost-related indicators, when the indicator value of the oil is greater than or equal to the indicator threshold, the corresponding oil indicator is defined as the oil indicator that caused the failure, and the traceability ends. When the index value of the oil is less than the index threshold, the attribute layer data of the oil is defuzzified starting from the second layer. The defuzzified values ​​are compared. For the attribute with the largest value, the failure source is traced along the corresponding attribute in the third layer branch. If the corresponding attribute in the third layer branch is all index data, the index data after unification of dimensions under the corresponding branch is compared, and the maximum value is set as the oil index that caused the failure. When the corresponding attribute has both indicator data and attribute data in the third-level branch, the attribute data is defuzzified and compared with the unified-dimensional indicator data in the corresponding branch. If the indicator data is the largest, the corresponding indicator is defined as the oil indicator that caused the failure. If the defuzzified attribute data is the largest, the source is traced back to the fourth-level branch. For the index data branch with attributes in the fourth layer, compare the index data with the unified dimensions under the corresponding branch, and set the maximum value as the oil index that caused the failure, thus completing the failure source tracing. For efficiency-related indicators, when the values ​​of various indicators of the oil are equal to or lower than the indicator threshold, the corresponding oil indicator is defined as the oil indicator that caused the failure, and the traceability ends.

Images