A maintenance analysis method for associating component failure causes by early warning patterns

Abstract

The application discloses a kind of maintenance analysis methods for the failure cause of component by early warning mode association, comprising: establishing the relationship between associated measuring point and early warning mode for thermal power plant equipment;Early warning mode and associated measuring point are associated with failure mode, failure cause in expert database, and failure consequence is quantified as inherent weight matrix μ of early warning mode;Establishing deviation degree calculation matrix, real-time monitoring updates the deviation degree of parameter;Establishing parameter grey correlation degree model, calculate the grey correlation degree between each parameter of equipment early warning mode;Establishing early warning risk model, calculate the early warning comprehensive risk value of early warning mode;According to early warning risk value, output maintenance strategy.The application adopts equipment operation monitoring to push maintenance strategy, combines real-time data model and eliminates data coupling by means of grey correlation degree, comprehensively judges failure early warning risk, and then pushes maintenance strategy.

Description

Technical Field

[0001] This invention relates to the field of condition-based maintenance of electrical equipment in thermal power plants, and specifically to a maintenance analysis method that associates the causes of component failures with early warning modes. Background Technology

[0002] Existing equipment inspection, maintenance, and early warning analysis in the market for thermal power plants remain at the equipment level, lacking a refined correlation between fault location and cause. Furthermore, there is a lack of further quantitative early warning risk assessment models after the early warning is issued, thus failing to further realize the function of assisting users in formulating maintenance tasks through early warning diagnosis.

[0003] The existing inspection, maintenance, and early warning analysis systems have the following shortcomings:

[0004] 1. The existing inspection and maintenance system remains at the equipment level, and equipment maintenance is carried out periodically. Users cannot conduct maintenance work by judging the actual fault status, resulting in a large number of over-maintenance and under-maintenance work.

[0005] 2. Existing early warning systems use equipment-level warning signals, which prevent users from accurately locating faulty equipment components in a short time based on the warning mode, increasing the response time for fault handling and resulting in ambiguous early warning positioning.

[0006] 3. After the early warning mode appears, there is no further quantitative fault analysis model to push maintenance strategies and form a closed-loop management system for refined information processing. Summary of the Invention

[0007] This invention provides a maintenance analysis method that associates component failure causes with early warning modes. It adopts equipment operation monitoring to push maintenance strategies, uses real-time data models and eliminates data coupling by using grey relational analysis, comprehensively evaluates the failure early warning risk, and then pushes maintenance strategies.

[0008] The present invention is a method that combines and analyzes the early warning mode of thermal power plant equipment with the fault mechanism of equipment components and real-time data, and solves the current problem of failing to combine the early warning mode of equipment with component fault management. The method systematically analyzes the application of the early warning mode of power generation equipment in equipment condition maintenance.

[0009] A maintenance analysis method that associates component failure causes with early warning modes includes the following steps:

[0010] 1.1) Establish relationships between associated monitoring points and early warning models for thermal power plant equipment;

[0011] 1.2) Associate the early warning mode and associated measurement points with the fault modes and fault causes in the expert database, and quantify the fault consequences as the inherent weight matrix μ of the early warning mode;

[0012] 1.3) Establish a deviation calculation matrix and monitor and update the deviation of parameters in real time;

[0013] 1.4) Establish a parameter grey relational model and calculate the grey relational degree between each parameter in the equipment early warning mode;

[0014] 1.5) Establish an early warning risk model and calculate the comprehensive early warning risk value of the early warning mode;

[0015] 1.6) Output maintenance strategies based on the early warning risk values.

[0016] In step 1.1), the thermal power plant equipment mentioned is a coal mill.

[0017] In step 1.2), the inherent weight matrix μ = [μ1 μ2 μ3…μ i …μ n ], where μ i The inherent weight of the fault mode causing the early warning mode to occur is defined by each parameter fault mode, where i represents the sequence number of the fault mode in the early warning mode.

[0018] In step 1.3), the deviation of the real-time monitoring and updating parameters is calculated using the following formula:

[0019]

[0020] Where i represents the sequence number of the fault mode in the warning mode, x i x is the measured value. max For the maximum value of the statistical parameter, For the average value, x min The greater the deviation between the monitored value and the normal value, the greater the risk of a warning. δ i (t) represents the deviation of each parameter from the change in time t.

[0021] In step 1.4), a parameter grey relational model is established to calculate the grey relational degree between various parameters of the equipment early warning mode, specifically including:

[0022] 1.4.1) Using real-time data, calculate the grey relational degree r between each associated measuring point. i Assuming there are n related measurement point parameters, and each parameter has m data points forming a data matrix, the following is a possible interpretation:

[0023]

[0024] 1.4.2) Using the reference data column x0, generate a dimensionless matrix to determine the minimum and maximum differences at both levels:

[0025] and Where k = 1, 2, 3...n;

[0026] 1.4.3) Calculate the correlation coefficient ζ i (k):

[0027]

[0028] Where ρ is 0.3 to 0.7, usually taken as 0.5, and k = 1, 2, 3...n.

[0029] 1.4.4) Calculate the grey relational degree r i :

[0030]

[0031] In step 1.5), an early warning risk model is established, and the comprehensive early warning risk value of the early warning mode is calculated, specifically including:

[0032] Based on the inherent weights μ of each parameter in the early warning mode i Dynamic deviation δ i (t) and grey relational degree r i Establish a comprehensive risk weighting index that combines static and dynamic factors. i :

[0033] w i =μ i δ i (t)r i , where i represents the sequence number of each parameter in the warning mode.

[0034] In reliability models, the failure rate is typically calculated using an exponential function. The failure rate is directly related to risk. The following uses an exponential model to calculate the early warning risk score H:

[0035]

[0036] Compared with the prior art, the present invention has the following advantages:

[0037] 1. This method proposes an early warning risk model that integrates the dynamic deviation of real-time data into the inherent weights of the mechanism model and eliminates the coupling between parameters. Maintenance work is carried out through early warning status inspection, reducing over- and under-maintenance work and saving maintenance costs.

[0038] 2. The equipment warning mode is associated with potentially faulty components and their causes – expert knowledge base. When a warning mode issues, the associated list allows users to narrow down the scope of troubleshooting before troubleshooting, reducing troubleshooting time.

[0039] 3. After the early warning mode is activated, it can not only display the related faulty components and their causes, but also push maintenance strategies or plans in real time through early warning risk calculation, forming a closed loop of early warning mode information data management.

[0040] 4. This invention integrates the dynamic deviation of real-time data into the inherent weights of the mechanism model, and combines it with the failure model to form a method for equipment failure early warning risk management analysis.

[0041] 5. This invention solves the problem of inaccurate positioning when the early warning mode occurs by directly associating the early warning mode with the faulty component, thus narrowing the scope of troubleshooting before the fault occurs.

[0042] 6. This invention combines inherent fault weights with real-time dynamic parameters, dynamically updates the weight index of the comprehensive risk assessment model in conjunction with the failure model, updates the fault warning risk value in real time, pushes maintenance strategies, and completes the closed loop of equipment warning management. Attached Figure Description

[0043] Figure 1 This is a flowchart of the solution in this invention;

[0044] Figure 2 This is the deviation variation curve in this invention;

[0045] Figure 3 This is a diagram of the fault early warning risk calculation model in this invention;

[0046] Figure 4 This is a flowchart of the maintenance strategy push process in this invention. Detailed Implementation

[0047] like Figure 1 As shown, the implementation steps are as follows:

[0048] a. Establish a relationship between associated monitoring points and early warning models for thermal power plant equipment.

[0049] Several variable measurement points of the equipment can trigger the same early warning mode. For example, common early warning modes of coal mills include: coal mill blockage, coal mill blockage, fire, and low output. The relevant measurement point indicators include: coal mill outlet temperature, coal mill current, coal mill primary air pressure, oil tank oil temperature, primary air flow, motor bearing temperature, and reducer thrust bearing temperature.

[0050] b. Associate the early warning mode and associated measurement points with the fault modes and causes in the expert database, and quantify the fault consequences as the inherent weight matrix μ of the early warning mode.

[0051] The standard library contains common equipment faults, their causes, and recommended maintenance measures, providing data support for equipment model building and analysis. It automatically retrieves equipment fault causes, symptoms, and consequences based on the expert standard library. For example, low output of a coal mill is a warning mode. Corresponding measurement points include coal feed rate, primary air flow rate, coal mill current, mill bowl pressure difference, and mill sealing air pressure difference. The corresponding fault modes in the expert library are: coal feed rate decrease, airflow decrease, abnormal coal mill current, abnormal mill bowl pressure difference, and abnormal mill sealing air pressure difference. The associated fault causes are shown in Table 1.

[0052] Table 1

[0053]

[0054] When each associated measuring point alarms, it triggers the occurrence of the corresponding fault mode, resulting in different fault consequences and influencing the occurrence of the warning mode to varying degrees. The inherent weight of each fault mode in causing the occurrence of the warning mode, obtained from the expert database, is denoted as μ. i 'i' represents the sequence number of the fault mode in the warning mode. The inherent weight matrix is ​​composed of μ = [μ1 μ2 μ3…μ]. i …μ n ].

[0055] The "low coal mill output" early warning mode is associated with 5 measuring points, and the obtained inherent weight matrix μ i = [0.7 0.6 0.5 0.4 0.2].

[0056] c. Establish a deviation calculation matrix and monitor the deviation of the updated parameters in real time.

[0057] Monitor associated measurement points, assuming the measurement point value is x. i The maximum value of the statistical parameter x max ,average value Minimum value x min The greater the deviation between the monitored value and the normal value, the greater the risk of a warning. The deviation is denoted as δ. i Then the deviation degree is related to x i The relationship curve is attached. Figure 2 As shown.

[0058] The specific formula for calculating the deviation of the actual monitoring point parameters is as follows:

[0059]

[0060] The deviation rate changes continuously as the equipment is monitored during operation. Based on historical data from measuring points corresponding to low output of the coal mill at a certain moment, the deviation rate was calculated using the formula, as shown in Table 2.

[0061] Table 2

[0062]

[0063] This yields the deviation matrix δ i = [0.64 0.48 0.91 0.52 0.72].

[0064] d. Establish a parameter grey relational model and calculate the grey relational degree between each parameter of the equipment early warning mode.

[0065] Grey relational analysis (GRA) is a multi-factor statistical analysis method. It uses sample data of various factors to describe the strength, magnitude, and order of the relationships between them. In equipment early warning modes, there may be coupling relationships between the various measuring points, so real-time data is needed to calculate the grey relational degree r between each measuring point. i Assuming there are n measurement points and parameters, and each parameter has m data points forming a data matrix, the following is a possible interpretation:

[0066]

[0067] By referencing data column x0, a dimensionless matrix is ​​generated to determine the minimum and maximum differences at both levels:

[0068] and Where k = 1, 2, 3...n;

[0069] Calculate the correlation coefficient:

[0070]

[0071] Where ρ is typically taken as 0.5, and k = 1, 2, 3...n.

[0072] Calculate the grey relational degree r i :

[0073]

[0074] The real-time monitoring data related to the "low coal mill output" early warning mode are used to construct a matrix:

[0075]

[0076] Based on the designed coal type, the ideal output of the coal mill is 70-80 t / h. A reference column matrix is ​​obtained by combining random numbers.

[0077] x0=[70.55 73.44 71.88 72.92 74.32 75.32 78.32 76.32...73.56 76.8679.32 78.54 77.44 76.43 75.44]

[0078] Find the dimensionless matrix:

[0079]

[0080] Thus, the minimum difference between the two levels is 0.02, and the maximum difference between the two levels is 106.17.

[0081] Substitute into the calculation of the correlation coefficient matrix:

[0082]

[0083] The final calculation yielded the grey relational matrix corresponding to "low coal mill output":

[0084]

[0085] e. Establish an early warning risk model and calculate the comprehensive early warning risk value of the early warning mode.

[0086] Based on the inherent weights μ of each parameter in the early warning mode i Dynamic deviation δ i (t) and grey relational degree r i Establish a comprehensive risk weighting index w that combines static and dynamic factors. i :

[0087] w i =μ i δ i (t)r i , where i represents the sequence number of each parameter in the warning mode.

[0088] Thus, μ i Taking into account the internal structure and failure mechanism of the equipment, δ i (t) Considering the real-time changes of the parameters, r i By taking into account the coupling relationship between parameters, this model is more accurate.

[0089] In reliability models, the failure rate is typically calculated using an exponential function. The failure rate is directly related to risk. The following uses an exponential model to calculate the warning score H:

[0090]

[0091] Therefore, based on the previous data, we continue to calculate the comprehensive risk weight index w for "low coal mill output". i :

[0092] w i =μ i δ i (t)r i =[0.4309 0.0988 0.3043 0.0884 0.0636]

[0093] Early warning risk score H:

[0094]

[0095] As shown in Table 3 below:

[0096] Table 3

[0097]

[0098] This establishes a complete fault early warning risk calculation model, see appendix. Figure 3 .

[0099] f. Output maintenance strategies based on the early warning risk values.

[0100] The comprehensive early warning risk value H is between 0 and 100%. Based on the early warning risk score of the early warning mode, an early warning strategy is pushed out when an alarm is triggered.

[0101] The maintenance strategy push process is attached. Figure 4 Different maintenance strategy push rules are divided according to the range of warning risk value. When 0≤H<20%, ​​the equipment is not considered for maintenance for the time being; when 20%≤H<40%, the system will prompt "Note, it is recommended to arrange maintenance at an opportune time"; when 40%≤H<60%, the system will prompt "Warning, it is recommended to arrange maintenance in the near future"; when 60%≤H≤100%, the system will prompt "It is recommended to perform maintenance immediately" and push a maintenance task.

[0102] The maintenance strategy push for the coal mill is shown in Table 4.

[0103] Table 4

[0104]

[0105] This method combines qualitative static parameters and quantitative dynamic parameters of faults, evaluates equipment based on dynamic early warning scores of the early warning mode, and proposes maintenance suggestions. Traditional scheduled maintenance methods involve fixed intervals between inspections, and maintenance may or may not occur during these intervals. In contrast, maintenance strategies based on the early warning risk model are more targeted, comprehensively considering the overall fault mechanism and real-time data. The model is more accurate; according to user statistics during model usage, the accuracy of maintenance strategies proposed for coal mills during operation can reach over 92%, reducing the phenomenon of over- and under-maintenance of equipment.

[0106] Ordinary early warning systems alarm at the device level and cannot quickly link to the critical fault location of the device. This method, after integrating the early warning model and the standard library device fault mechanism model, adds the association with the device fault mode, fault cause, and faulty component. This enables the early warning mode to accurately and quickly locate the faulty component and the possible fault cause, and push the fault prompt list and maintenance task in real time. At the same time, it dynamically updates the early warning weight index according to the accumulation of equipment data and changes in the operating environment, completing the closed loop of early warning data management.

[0107] This system can improve enterprises' comprehensive management of equipment early warning data and faults, enhance response speed to equipment early warnings, and push maintenance tasks in real time. It provides a basis for equipment maintenance planning, reduces over- and under-maintenance in scheduled maintenance, and optimizes equipment maintenance strategies based on dynamic risk assessments of power plant equipment early warnings. This extends the maintenance interval for over-maintenance equipment, correspondingly reducing manpower and spare parts investment. Taking a coal mill as an example, the cost of external maintenance personnel in a power plant is approximately 200 yuan per man-hour. A minor repair costs 30,000 yuan in manpower and 38,000 yuan in spare parts. A major overhaul costs 90,000 yuan in manpower and 905,000 yuan in spare parts. The annual maintenance cost of a coal mill, excluding internal manpower, is around 530,000 yuan. Using this early warning risk model to dynamically push maintenance strategies and conduct maintenance work based on the coal mill's equipment status can shorten the annual maintenance time, reducing the annual maintenance cost to around 460,000 yuan, a saving of 13%.

Claims

1. A maintenance analysis method that correlates component failure causes through early warning modes, characterized in that, Includes the following steps: 1.1) Establish relationships between associated monitoring points and early warning models for thermal power plant equipment; 1.2) Associate the early warning mode and associated measurement points with the fault modes and fault causes in the expert database, and quantify the fault consequences as the inherent weight matrix of the early warning mode. ; The inherent weight matrix ,in, The inherent weight of the early warning mode caused by each parameter's fault mode is i, where i represents the sequence number of the fault mode in the early warning mode. 1.3) Establish a deviation calculation matrix and monitor and update the deviation of parameters in real time; The dynamic deviation of the real-time monitoring and updating parameters is calculated using the following formula: Where i represents the sequence number of the fault mode in the warning mode, For the measured point value, For the maximum value of the statistical parameter, For average value, The greater the deviation between the monitored value and the normal value, the greater the risk of a warning. The deviation of each parameter from the change over time t; 1.4) Establish a parameter grey relational model and calculate the grey relational degree between each parameter of the equipment early warning mode; 1.5) Establish an early warning risk model and calculate the comprehensive early warning risk value of the early warning mode, specifically including: Based on the inherent weights of each parameter in the early warning mode Dynamic deviation and grey relational degree Calculate the comprehensive risk weight index that combines static and dynamic factors. : , where i represents the sequence number of each parameter in the warning mode; In reliability models, the failure rate is typically calculated using an exponential function. The failure rate is directly related to risk. Below, we will use an exponential model to calculate the early warning risk score. : ； 1.6) Output maintenance strategies based on the early warning risk values.

2. The maintenance analysis method for associating component failure causes through early warning mode according to claim 1, characterized in that, In step 1.1), the thermal power plant equipment mentioned is a coal mill.

3. The maintenance analysis method for associating component failure causes through early warning mode according to claim 1, characterized in that, In step 1.4), a parameter grey relational model is established to calculate the grey relational degree between various parameters of the equipment early warning mode, specifically including: 1.4.1) Using real-time data, calculate the grey relational degree between each associated measuring point. Assuming there are n related measurement point parameters, and each parameter has m data points forming a data matrix, the following is a possible interpretation: 1.4.2) By referencing data columns Generate a dimensionless matrix and determine the minimum and maximum differences at both levels: and Where k = 1, 2, 3...n; 1.4.3) Calculate the correlation coefficient : Where ρ is 0.3~0.7, k=1,2,3...n; 1.4.4) Calculate the grey relational degree : 。

Images