Simulation sample generation method for device fault mechanisms

Abstract

A simulation sample generation method for device fault mechanisms. The method comprises the steps of: step I, analyzing device fault mechanisms; step II, determining fault parameters: determining fault parameters on the basis of the device fault mechanisms; step III, in view of the knowledge in the field of hydroelectric-device fault detection, determining corresponding credibility algorithms for different device fault symptom modes on the basis of the fault parameters; step IV, performing tuning on the basis of the formulas of the credibility algorithms to obtain templates corresponding to the different fault symptom modes; and step V, generating simulated data on the basis of the corresponding templates. In the present invention, data that meets device fault diagnosis conditions and corresponds to different symptom modes is generated on the basis of an industrial mechanism, so that simulation samples can reflect important values in terms of monitoring, optimization and effect prediction.

Description

A method for generating simulation samples based on equipment failure mechanisms Technical Field

[0001] The present invention belongs to the field of industrial Internet platforms, and in particular relates to a method for generating simulation samples for equipment failure mechanisms. Background Art

[0002] Currently, data analysis research requires the use of large amounts of real data to ensure that the analysis results approximate actual results. However, in practice, the amount of real data is insufficient to meet the needs of data analysis. Therefore, data simulation is often used during data analysis research. This involves using simulated samples to simulate the data and then performing analysis based on the resulting simulated data. The significance of simulated samples lies in the fact that actual equipment operating data does not meet the data input requirements for fault diagnosis. Simulated samples are often used to generate simulated data for fault diagnosis simulations, providing sample services for advanced applications, data analysis development tools, and fault case libraries. Once generated, simulated samples exist in a separate simulated time series library. Modifications and edits to the data are isolated from the production library and do not affect the actual data.

[0003] Existing simulation data generation methods typically simulate data based on real historical data to generate large amounts of simulated data corresponding to the historical data. However, in reality, real data may contain outliers, especially abnormal maximum and minimum values, which can lead to an excessively large range of corresponding data values, affecting efficiency and accuracy. Therefore, post-simulation data editing and modification are cumbersome to meet data requirements.

[0004] Summary of the Invention

[0005] In view of the technical problems existing in the background technology, the present invention provides a simulation sample generation method for equipment failure mechanism, which generates data corresponding to different symptom patterns based on industrial mechanisms and meets the equipment fault diagnosis conditions, enabling the simulation samples to reflect important value in monitoring, optimization, and effect prediction.

[0006] In order to solve the above technical problems, the present invention adopts the following technical solutions:

[0007] A method for generating simulation samples for equipment failure mechanisms, including the following steps:

[0008] Step 1: Analyze the equipment failure mechanism;

[0009] Step 2: Determine the fault parameters: Determine the fault parameters based on the equipment failure mechanism;

[0010] Step 3: Combined with the knowledge of hydropower equipment fault detection, determine the corresponding credibility algorithm for different equipment fault symptom patterns based on the fault parameters;

[0011] Step 4: Optimize according to the credibility algorithm formula to obtain templates corresponding to different fault symptom patterns;

[0012] Step 5: Generate simulation data according to the corresponding templates.

[0013] Preferably, in Step 1, the fault mechanisms include the following categories:

[0014] The first category: large upper guide swing;

[0015] The second category: vibration based on the rotation frequency;

[0016] The third category: vibration linearly increases with the square of the rotational speed;

[0017] The fourth category: vibration is independent of load.

[0018] Preferably, in Step 3, 1) when the fault symptom is the first category, set the overall frequency amplitude ugt of the upper guide swing and the channel danger value ugtd. The credibility calculation formula (1) is: cf = min(1.0, 1.2 * ugt / ugtd) (1);

[0019] According to the credibility calculation formula (1), the closer the cf value is to 1, the more similar the symptom is; so 1.2 * ugt / ugtd ≥ 1.0, and by reverse deduction, the generation formula one is obtained: ugt / ugtd > 0.834 (one);

[0020] The symptom can be satisfied;

[0021] 2) When the fault symptom is the second category, if the rotation frequency fyp1 of the vibration swing channel is greater than the reference value of the frequency band, based on the overall frequency amplitude fypp, the credibility calculation formula (2) is: cf = min(1.0, 1.2 * fyp1 / fypp) (2);

[0022] According to the credibility calculation formula (2), the closer the cf value is to 1, the more similar the symptom is; so 1.2 * fyp1 / fypp ≥ 1.0, and by reverse deduction, the generation formula two is obtained: fyp1 / fypp > 0.834 (two);

[0023] The symptom can be satisfied;

[0024] 3) When the fault symptom is the third category, during the speed-up process of the unit, take 40 - 60% of the rated speed as speed1, 80 - 100% of the rated speed as speed2, the vibration swing amplitude at speed1 is fypp1, and the vibration swing amplitude at speed2 is fypp2. Calculate the changes in rotational speed and vibration swing as speedb and fyppb respectively;

[0025] Then speedb=(speed2 / speed1) 2 ,

[0026] fyppb=fypp2 / fypp1;

[0027] Credibility formula (3) is as follows:

[0028] If fyppb / speedb>0.75 and fyppb / speedb<1.5, then cf=1, otherwise 0 (3);

[0029] According to the credibility formula (3), the reverse deduction produces the formula 3: 1.5>((fypp2 / fypp1) / (speed2 / speed1) 2 >0.75 (three);

[0030] To satisfy the symptoms;

[0031] fypp2 is the swing 1X amplitude when the unit speed is 80-100% of the rated speed;

[0032] fypp1 is the swing 1X amplitude when the unit speed is 40-60% of the rated speed;

[0033] speed2 is 80-100% of rated speed;

[0034] speed1 is 40-60% of rated speed;

[0035] 4) When the fault symptom is the fourth type, the full-frequency amplitude of the vibration swing under the load less than 10% of the rated load within 3 hours is fyppLP, and the full-frequency amplitude of the vibration swing under the load greater than 90% of the rated load is fyppHP. The vibration change is fypb=fyppHP-fyppLP, and the amplitude stability condition is fypstatic. If fypb>0, the reliability calculation formula (4) is: cf=1-min(1.0,0.25*fypb / fypstatic); (4);

[0036] According to the credibility calculation formula (4), the generation formula 4 is derived: 1>(fyppHP-fyppLP) / fypstatic>0 (IV);

[0037] The signs are satisfied.

[0038] Preferably, in step 4, 1) when the fault symptom is of the first type, the adjusted credibility calculation formula is: ugt / ugtd>0.834;

[0039] 2) When the fault symptom is the second type, the adjusted reliability calculation formula is: 1X amplitude / peak-to-peak swing>0.85;

[0040] 3) When the fault symptom is the third type, the adjusted reliability calculation formula is: 1.5>((fypp2 / fypp1) / (speed2 / speed1) 2 >0.75;

[0041] 4) When the fault symptom is the fourth category, the adjusted reliability calculation formula is: 1>(fyppHP-fyppLP) / fypstatic>0.1.

[0042] Preferably, in step five, a time range is selected through a symptom pattern formula template, and simulation is performed based on real data to generate simulated data that meets the symptom, that is, simulated sample generation.

[0043] The present invention can achieve the following beneficial effects:

[0044] Because actual hydropower equipment operating data sometimes doesn't meet the data input requirements for fault diagnosis, simulated samples are often used in fault diagnosis simulations. This invention, based on industrial mechanisms, generates data corresponding to different symptom patterns that meet the equipment fault diagnosis criteria. This enables simulated samples to demonstrate significant value in monitoring, optimization, and effect prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] The present invention will be further described below with reference to the accompanying drawings and embodiments:

[0046] FIG1 is a flow chart of the present invention. DETAILED DESCRIPTION

[0047] Example 1:

[0048] The preferred solution is shown in Figure 1. It is a simulation sample generation method for equipment failure mechanisms. It can generate data that meets the requirements based on industrial mechanisms and provide data set services for other modules. The specific steps are as follows:

[0049] Step 1: Analyze the equipment failure mechanism: First, you need to understand the working principle of the equipment and the possible failure mechanisms. For example, for a hydro-turbine generator, possible failure mechanisms include large upper guide swing.

[0050] Step 2: Determine fault parameters: Determine fault parameters based on the equipment failure mechanism. For example, a generator upper conductor swing fault may be caused by excessive turbine vibration, and the fault parameters include peak-to-peak swing value.

[0051] Step 3: Based on the fault parameters, analyze the hydropower equipment fault field and combine the equipment fault symptoms with a specific credibility algorithm. For example:

[0052] 1) Large upper guide swing: Let the overall frequency amplitude ugt (peak-to-peak value) of the upper guide swing, the channel danger value ugtd (upper limit value of peak-to-peak value), and the credibility calculation formula (1) be: cf = min(1.0, 1.2 * ugt / ugtd) (1);

[0053] According to the credibility calculation formula (1), the closer the cf value is to 1, the more similar the symptoms are. Therefore, 1.2 * ugt / ugtd ≥ 1.0, and by reverse deduction, formula one is generated: ugt / ugtd > 0.834 (one);

[0054] The symptoms can be satisfied;

[0055] 2) Vibration is based on the rotation frequency (i.e., vibration is mainly based on the rotation frequency): If the rotation frequency fyp1 (1X amplitude) of the vibration swing channel is greater than the reference value of the frequency band, based on the overall frequency amplitude fypp (peak-to-peak value), the credibility calculation formula (2) is: cf = min(1.0, 1.2 * fyp1 / fypp) (2);

[0056] According to the credibility calculation formula (2), the closer the cf value is to 1, the more similar the symptoms are. Therefore, 1.2 * fyp1 / fypp ≥ 1.0, and by reverse deduction, formula two is generated: 1X amplitude / swing peak-to-peak value > 0.834 (two);

[0057] The symptoms can be satisfied.

[0058] 3) Vibration increases linearly with the square of the rotational speed:

[0059] During the speed-up process of the unit, take the vibration swing amplitudes at 40 - 60% of the rated speed (set as speed1) and 80 - 100% of the rated speed (set as speed2) as fypp1 and fypp2 respectively, and calculate the changes in rotational speed and vibration swing as speedb and fyppb respectively,

[0060] Then speedb = (speed2 / speed1) 2 ,

[0061] LThe credibility formula (3) is as follows:

[0062] fyppb = fypp2 / fypp1, if fyppb / speedb > 0.75 and fyppb / speedb < 1.5, then cf = 1, otherwise it is 0 (3);

[0063] According to the credibility formula (3), by reverse deduction, formula three is generated: 1.5 > ((fypp2 / fypp1) / (speed2 / speed1)2 >0.75 (three);

[0064] To satisfy the symptoms;

[0065] fypp2 is the swing 1X amplitude when the unit speed is 80-100% of the rated speed;

[0066] fypp1 is the swing 1X amplitude when the unit speed is 40-60% of the rated speed;

[0067] speed2 is 80-100% of rated speed;

[0068] speed1 is 40-60% of the rated speed.

[0069] 4) Vibration is independent of load: Assume that the full-frequency amplitude of the vibration swing under the load (active power) less than 10% of the rated load within 3 hours is fyppLP, and the full-frequency amplitude of the vibration swing under the load greater than 90% of the rated load is fyppHP. The vibration change is fypb = fyppHP-fyppLP, and the amplitude stability condition is fypstatic. If fypb>0, the reliability calculation formula (4) is: cf = 1-min(1.0,0.25*fypb / fypstatic) (4);

[0070] According to the credibility formula (4), the formula 4 is derived in reverse: 1>(fyppHP-fyppLP) / fypstatic>0 (IV);

[0071] The signs are satisfied.

[0072] Step 4: Repeatedly test and debug by generating formulas 1 to 4, and finally use the tuned formula as a template for this symptom pattern.

[0073] During the debugging process, the influence of some critical values ​​is avoided, and the upper and lower limits of the range are averaged, and then the formula is optimized so that the formula can fully meet the symptoms.

[0074] 1) The upper guide swing is large. The formula before adjustment is: the peak-to-peak value of the swing / the upper limit of the peak-to-peak value of the swing>0.834. The formula after adjustment is: the peak-to-peak value of the swing / the upper limit of the peak-to-peak value of the swing>0.85.

[0075] 2) Vibration is based on rotational frequency. The formula before adjustment is: 1X amplitude / peak-to-peak value of swing> 0.834. The formula after adjustment is: 1X amplitude / peak-to-peak value of swing> 0.85.

[0076] 3) Vibration increases linearly with the square of the speed. The formula before adjustment is: 1.5>((fypp2 / fypp1) / (speed2 / speed1) 2>0.75,

[0077] Adjusted formula: 1.5>((fypp2 / fypp1) / (speed2 / speed1) 2 >0.75.

[0078] 4) Vibration is independent of load. The formula before adjustment is: 1>(fyppHP-fyppLP) / fypstatic>0. The formula after adjustment is: 1>(fyppHP-fyppLP) / fypstatic>0.1.

[0079] Step 5: Use the symptom pattern formula template to select a time range, use real data as the basis, perform simulation, and generate simulated data that meets the symptoms, that is, simulated sample generation.

[0080] The generation of simulated samples requires consideration of diversity and randomness to ensure representativeness and reliability. Furthermore, parameter selection and template creation must be tailored to actual conditions to ensure the generated samples reflect the actual situation. This invention supports the generation of simulated samples based on equipment failure mechanisms through symptom patterns. Symptom patterns and fault symptoms can be understood as phenomena, such as large upper guide oscillation or large water guide oscillation. These are generally quantitatively defined by ensuring that the values ​​at the relevant measurement points fall within a defined numerical range.

[0081] As shown in Table 1, the simulation samples of this method support the generation of data for four symptom modes based on industrial mechanisms:

[0082] Table 1 Data generated by four symptom modes

[0083] Table 1 includes the settings of three items: symptom name, related measurement points, and data formula. Its beneficial effects are:

[0084] 1. Clear fault characteristics and monitoring direction: The symptom name directly points out the specific fault type that the equipment may encounter, such as "large upper guide swing" and "vibration mainly in rotation frequency". This helps maintenance personnel quickly focus on the problem and clarify the direction of monitoring and diagnosis;

[0085] 2. Guidance on measurement point selection and data collection: The relevant measurement points list the specific sensors or measurement locations that need to be monitored to identify and analyze specific fault symptoms, such as swing sensors and vibration sensors. This is of great significance for the rational layout of the monitoring network and ensuring comprehensive and accurate data collection, which can reduce monitoring blind spots and improve the accuracy of fault diagnosis;

[0086] 3. Standardized and Formulated Fault Analysis: The provision of data formulas not only unifies the fault analysis methodology, but also makes the fault identification process more standardized and automated. These formulas can quickly calculate indicators of specific fault symptoms and quantify the degree of fault;

[0087] 4. Promote the construction of fault prediction models: Based on the data formulas in the table, they can be directly applied to build or optimize fault prediction models, and use historical and real-time data to predict the possibility of future faults.

[0088] The above embodiments are merely preferred technical solutions of the present invention and should not be construed as limiting the present invention. The scope of protection of the present invention shall be the technical solutions set forth in the claims, including equivalent alternatives to the technical features of the technical solutions set forth in the claims. In other words, equivalent alternatives and improvements within this scope are also within the scope of protection of the present invention.

Claims

1. A method for generating simulation samples for equipment failure mechanisms, characterized in that It includes the following steps: Step 1: Analyze the equipment failure mechanism; the failure mechanism includes four categories: The first category: large upper guide swing; the second category: vibration is based on the rotation frequency; The third category: vibration shows a linear increase with the square of the rotational speed; The fourth category: vibration is independent of the load; Step 2: Determine the failure parameters: According to the equipment failure mechanism, determine the failure parameters; Step 3: Combining the knowledge in the field of hydropower equipment failure detection, determine the corresponding credibility algorithm for different equipment failure symptom patterns according to the failure parameters; Step 4: Optimize according to the credibility algorithm formula to obtain templates corresponding to different failure symptom patterns; Step 5: Generate simulation data according to the corresponding templates.

2. The method for generating simulation samples for equipment failure mechanism according to claim 1, characterized in that: In Step 3, 1) When the failure symptom is the first category, let the overall amplitude ugt of the upper guide swing and the channel danger value ugtd be set, and the credibility calculation formula (1) is: cf = min(1.0, 1.2 * ugt / ugtd) (1); According to the credibility calculation formula (1), the closer the cf value is to 1, the more similar the symptom is; therefore, 1.2 * ugt / ugtd ≥ 1.0, and by reverse deduction, formula one is generated: ugt / ugtd > 0.834 (1); That is, the symptom can be satisfied; 2) When the failure symptom is the second category, if the rotation frequency fyp1 of the vibration swing channel is greater than the reference value of the frequency band, based on the overall amplitude fypp, the credibility calculation formula (2) is: cf = min(1.0, 1.2 * fyp1 / fypp) (2); According to the credibility calculation formula (2), the closer the cf value is to 1, the more similar the symptom is; therefore, 1.2 * fyp1 / fypp ≥ 1.0, and by reverse deduction, formula two is generated: fyp1 / fypp > 0.834 (2); That is, the symptom can be satisfied; 3) When the failure symptom is the third category, during the unit speed-up process, take 40 - 60% of the rated speed as speed1, 80 - 100% of the rated speed as speed2, the vibration swing amplitude at speed1 is fypp1, and the vibration swing amplitude at speed2 is fypp2. Calculate the changes in rotational speed and vibration swing as speedb and fyppb respectively; Then speedb = (speed2 / speed1) 2 , fyppb = fypp2 / fypp1; The credibility formula (3) is as follows: If fyppb / speedb > 0.75 and fyppb / speedb < 1.5, then cf = 1, otherwise it is 0 (3); According to the credibility formula (3), by reverse deduction, formula three is generated: 1.5 > ((fypp2 / fypp1) / (speed2 / speed1) 2 > 0.75 (III); That is, the symptom can be satisfied; fypp2 is the 1X amplitude of the swing when the unit speed is 80 - 100% of the rated speed; fypp1 is the 1X amplitude of the swing when the unit speed is 40 - 60% of the rated speed; speed2 is 80 - 100% of the rated speed; speed1 is 40 - 60% of the rated speed; 4) When the fault symptom is the fourth type, the full-frequency amplitude of the vibration swing under the load less than 10% of the rated load within 3 hours is fyppLP, and the full-frequency amplitude of the vibration swing under the load greater than 90% of the rated load is fyppHP, the vibration change is fypb=fyppHP-fyppLP, and the amplitude stability condition is fypstatic; if fypb>0, the credibility calculation formula (4) is: cf=1-min(1.0,0.25*fypb / fypstatic) (4); According to the credibility calculation formula (4), the formula 4 is derived: 1>(fyppHP-fyppLP) / fypstatic>0 (IV); The signs can be satisfied.

3. The method for generating simulation samples for equipment failure mechanism according to claim 2, characterized in that: In step 4, 1) when the fault symptom is the first type, the adjusted reliability calculation formula is: ugt / ugtd>0.834; 2) When the fault symptom is the second type, the adjusted reliability calculation formula is: 1X amplitude / peak-to-peak swing value>0.85; 3) When the fault symptom is the third type, the adjusted reliability calculation formula is: 1.5>((fypp2 / fypp1) / (speed2 / speed1) 2 >0.75; 4) When the fault symptom is of the fourth type, the adjusted reliability calculation formula is: 1>(fyppHP-fyppLP) / fypstatic>0.

1.

4. The method for generating simulation samples for equipment failure mechanism according to claim 3, characterized in that: In step five, a time range is selected through the symptom pattern formula template, and simulation is performed based on real data to generate simulated data that meets the symptom, that is, simulated sample generation.

Images