Digital twin health management method for turbine component testing

Abstract

The present application belongs to the technical field of operation and maintenance of turbine component testing equipment, and particularly relates to a digital twin health management method for turbine component testing. The specific technical scheme is as follows: a plurality of sensors are arranged, a water power testing equipment related performance parameter database is established, and a digital twin health management method of the water power testing equipment is constructed based on a physical operation mechanism of a turbine-water power testing system and a water power testing equipment performance digital twin model of operation data. The method can realize accurate real-time monitoring and overrun early warning of key operation parameters of the water power testing equipment, improve the safety of the testing system, and reduce the operation and maintenance cost of the system.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of operation and maintenance of turbine component test equipment, and particularly relates to a digital twin health management method for turbine component testing. BACKGROUND

[0002] The hydraulic dynamometer equipment is a device that absorbs and transmits the output power of a power machine by using the friction torque of water on a rotating rotor, and is mainly used to measure the shaft power characteristics of machines such as generators, motors, internal combustion engines, gas turbines and pumps. Compared with other forms of dynamometer equipment, the hydraulic dynamometer equipment has the advantages of strong unit mass absorption function, small moment of inertia, wide rotational speed and power measurement range, convenient adjustment, high stability and the like, and is widely used in the test experiments of aero-engines and gas turbines.

[0003] Figure 1 is a basic structure diagram of the turbine hydraulic dynamometer equipment, and the basic working principle is as follows: the dynamometer disc is fixed on the rotating shaft and rotates together with the rotating shaft in the shell 3, forming the rotor 1 of the dynamometer equipment. The rotor 1 is supported by bearings in the shell 3, and the shell 3 is supported by outer bearings on the dynamometer equipment base 5, which can freely swing around the axis. Water flows into the inner cavity of the shell 3 through the water inlet valve. When the flywheel of the internal combustion engine drives the rotor 1 to rotate in the shell, the water also rotates together due to the friction between the rotating disc and the water. Under the action of centrifugal force, the water is thrown to the inner wall of the shell 3, forming an annular water ring and increasing its momentum moment. At the same time, the shell 3 swings around the axis after being impacted by the water. The water impacting the inner wall of the shell 3 is subjected to the friction resistance of the inner wall, and the speed is reduced, and the kinetic energy is paid. Under the action of water pressure, the water flows to the center of the shell 3, forming an annular vortex water ring, and finally the temperature of the water is increased. Due to the friction between the water ring and the inner wall of the shell 3, the rotation (swing) speed of the shell 3 is slower than that of the rotor 1, so the water ring resists the rotation of the rotor 1 and generates a resistance torque. The resistance torque acts directly on the flywheel of the internal combustion engine through the shaft coupling 2, which is the load applied to the internal combustion engine. Or, the water absorbs the power of the internal combustion engine. This load can be controlled by adjusting the water inlet and outlet. The more the water ring, the greater the resistance torque, and the more power is absorbed.

[0004] Most of the existing large hydraulic dynamometer equipment is imported from abroad, and its internal structure is complex, and the detailed operation mechanism is not clear, which brings great difficulty to daily maintenance and repair. In addition, with the long-time operation of the hydraulic dynamometer equipment, due to the wear and aging of the mechanical structure, the performance of the equipment will inevitably decline, and the current characteristic curve will deviate greatly from that at the time of leaving the factory. If left unchecked, it may interfere with normal experimental tests.

[0005] At present, when the turbine test operation and state monitoring are carried out in China, the artificial experience diagnosis method is mainly relied on, that is, the experimental personnel rely on rich practical experience and theoretical knowledge to identify the working state of the equipment by listening to the abnormal sound in the experimental process and observing the experimental monitoring data. The artificial experience judgment method greatly depends on the technical level of the staff, and since the technical personnel are uneven, the long-term reliability of the judgment is difficult to guarantee. The artificial experience judgment method is difficult to summarize and reproduce to provide guidance for future experiments, and many times only the equipment failure is found, but the specific reasons leading to the failure cannot be accurately summarized.

[0006] Therefore, if a digital twin health management method for turbine component testing can be provided, it will have great application value. SUMMARY

[0007] To solve the above technical problems, the purpose of the present application is to provide a digital twin health management method and platform for turbine component testing.

[0008] To achieve the above-mentioned purpose of the application, the technical scheme adopted by the present application is: a digital twin health management method for turbine component testing, comprising the steps of:

[0009] S01, arranging a plurality of sensors on a turbine hydraulic dynamometer equipment to monitor the performance parameters of the hydraulic dynamometer equipment in real time;

[0010] S02, establishing a hydraulic dynamometer equipment performance parameter database, storing the sensor monitoring data into the database, converting the voltage signal of the sensor into the engineering value data corresponding to the performance parameter according to the preset conversion relationship, and storing the engineering value data into the database after screening processing;

[0011] S03, according to the input requirements of the hydraulic dynamometer equipment performance prediction model, extracting test data according to the storage frequency of the sensor monitoring data in the step S02;

[0012] S04, inputting the test data extracted in the step S03 into the hydraulic dynamometer equipment performance prediction model to calculate the performance parameters of the hydraulic dynamometer equipment;

[0013] S05, calculating the relative error between the performance parameter prediction value in the step S04 and the performance parameter acquisition value acquired by the sensor;

[0014] S06, judging the running state of the hydraulic dynamometer equipment according to the performance parameter acquisition value, the performance parameter prediction value and the prediction accuracy of the hydraulic dynamometer equipment.

[0015] Preferably, the sensor in the step S01 comprises a pressure sensor, a temperature sensor, a flow sensor, a speed sensor, a vibration sensor; and / or, the performance parameter comprises a rotating speed, a torque, an inlet water pressure, an oil supply pressure, an inlet water valve opening, a drain valve opening, a front bearing oil flow, a rear bearing oil flow, an inlet water temperature, a drain water temperature, a front bearing temperature, a rear bearing temperature, a front bearing vibration parameter, and a rear bearing vibration parameter.

[0016] Preferably, the engineering value data screening processing method in the step S02 is as follows: data within a rotating speed of 1000 rpm is removed as invalid data; data with an inlet and drain valve position lower than 1% is removed as interference data; data with a front and rear bearing temperature higher than 70℃ is removed from the training data as abnormal data for model accuracy verification in the test link.

[0017] Preferably, the hydraulic power test equipment running state judgment method in the step S06 is as follows,

[0018] S061, if the performance parameter collection value and the performance parameter prediction value exceed the limit value, or the prediction accuracy exceeds the first threshold value, it is determined whether to continue testing according to the abnormal situation;

[0019] If the test is continued, the test environment is changed and the step S01 is entered to continue the test, the abnormal data at the current time is removed and stored in the abnormal database table in the database;

[0020] If the test is not continued, the test is ended, and the performance prediction model is updated and corrected according to all the health data in the test process;

[0021] S062, if the performance parameter collection value and the performance parameter prediction value do not exceed the limit value, and the prediction accuracy does not exceed the first threshold value, it is determined whether to continue testing according to the predetermined test plan;

[0022] If the test is continued, the performance parameter prediction value and the prediction accuracy at the current time are stored in the prediction tracking result table in the database, and the step S01 is entered to continue the data collection and performance tracking prediction at the next time;

[0023] If the current time is the final test time, the test is terminated, and the performance prediction model is updated and corrected according to all the health data in the test process.

[0024] Preferably, it further comprises a step S07 of performing migration learning on the hydraulic equipment performance degradation detection, component performance degradation detection, and degradation characteristics in the performance prediction model.

[0025] Preferably, the step S07 comprises,

[0026] S071, using the current time test cycle data, extracting the current performance state characteristics according to the hydraulic dynamometer performance digital twin model, training the current test cycle performance model M';

[0027] S072, digital test of M' on original data: input state of original state into M', get performance feedback result of M' to original state;

[0028] S073, calculate component degradation factor: assume that the mean value of component core performance parameter P of hydraulic test equipment in original state is The prediction result of M' to original state P is Then the degradation factor of parameter P is

[0029] S074, calculate degradation factor difference: the alpha of each component performance parameter of this test cycle and the last time p Difference;

[0030] S075, judge whether it is necessary to migrate and train the component network layer:

[0031] If the alpha p Difference is within the second threshold range, it is considered that the performance of each component has not degraded, and the current performance prediction model can continue to be used;

[0032] If the alpha p Difference exceeds the second threshold, the components exceeding the second threshold are defined as performance degradation components, and only the network layer of the performance degradation components is migrated and trained in the subsequent performance prediction model updating process.

[0033] Correspondingly: a management platform of a digital twin health management method for turbine component testing, comprising an acquisition module, a database, a data processing module, a performance prediction module, an analysis module, the acquisition module is in communication connection with the data processing module, the performance prediction module is in communication connection with the analysis module, the data processing module, the performance prediction module and the analysis module are connected with the database;

[0034] The acquisition module acquires hydraulic dynamometer performance parameters and transmits them to the data processing module;

[0035] The data processing module performs screening processing on the performance parameter acquisition values, and stores them in the abnormal data table and the health data table in the database according to whether the data is abnormal;

[0036] The performance prediction module extracts test data from the database and calculates the performance parameters of the hydraulic dynamometer;

[0037] The analysis module calculates the relative error between the performance parameter prediction value and the performance parameter acquisition value collected by the sensor, and judges the running state of the hydraulic dynamometer equipment.

[0038] Preferably, the acquisition module comprises a pressure sensor, a temperature sensor, a flow sensor, a speed sensor, a vibration sensor.

[0039] Preferably, the performance parameters include rotational speed, torque, water inlet pressure, lubricating oil supply pressure, water inlet valve opening, water outlet valve opening, front bearing lubricating oil flow, rear bearing lubricating oil flow, water inlet temperature, water outlet temperature, front bearing temperature, rear bearing temperature, front bearing vibration parameter and rear bearing vibration parameter.

[0040] Correspondingly, an electronic device comprises:

[0041] One or more processors;

[0042] A storage device for storing one or more programs;

[0043] When the one or more programs are executed by the one or more processors, the one or more processors implement a digital twin health management method for turbine component testing.

[0044] Correspondingly, a computer readable medium stores a computer program, which, when executed by a processor, implements a digital twin health management method for turbine component testing.

[0045] Compared with the prior art, the present application has the following beneficial effects:

[0046] The present application sets up a plurality of sensors, establishes a hydraulic dynamometer related performance parameter database, and based on the physical operation mechanism of the turbine-hydraulic dynamometer system and the hydraulic dynamometer performance digital twin model of the operation data, a digital twin health management method of the hydraulic dynamometer is constructed. The method can realize accurate real-time monitoring and overrun early warning of the key operation parameters of the hydraulic dynamometer, improve the safety of the test system, and at the same time reduce the operation and maintenance cost of the system. At the same time, the performance degradation factor calculation of the core components of the turbine-hydraulic test system can be realized, which provides guidance for the performance state evaluation and maintenance instruction of the components. BRIEF DESCRIPTION OF DRAWINGS

[0047] Figure 1 The figure is a schematic diagram of the turbine hydraulic dynamometer structure of the present application;

[0048] Figure 2 The figure is a digital twin health management platform workflow diagram of the present application;

[0049] Figure 3 The figure is a performance prediction model updating process flowchart of the present application;

[0050] Figure 4 This is the data flow structure of the digital twin health management platform of the present invention;

[0051] Figure 5 This is a flowchart of the digital twin performance prediction model for turbine component testing according to the present invention;

[0052] Figure 6 This invention introduces a flowchart of a digital twin performance prediction model for the turbine system coupling network layer and the dynamometer system coupling network layer.

[0053] Figure 7 A flowchart of the digital twin performance prediction model for introducing the connection correction network layer is provided for this invention.

[0054] Figure 8 This is a schematic diagram of the bidirectional structure of the physical information network layer of the present invention;

[0055] Figure 9 This is a flowchart of the digital twin performance prediction model for turbine component testing based on the cumulative effect of testing, as described in this invention.

[0056] Figure 10 This is a diagram of the internal structure of the cumulative effect network layer in this invention;

[0057] Figure 11 This is a flow chart of the digital twin health management platform of the present invention.

[0058] The attached diagram is labeled as follows: Rotor 1, Coupling 2, Housing 3, Main Shaft 4, Base 5, Acquisition Module 6, Database 7, Data Processing Module 8, Performance Prediction Module 9, Analysis Module 10. Detailed Implementation

[0059] 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 embodiments of the present invention, and not all embodiments. Unless otherwise specified, the technical means used in the embodiments are conventional means well known to those skilled in the art.

[0060] like Figure 2 , 3 As shown, this invention discloses a digital twin health management method for turbine component testing, which involves collecting, filtering, and storing test parameters of a turbine hydraulic dynamometer; real-time tracking, monitoring, and predicting key performance parameters of the hydraulic dynamometer; calculating the performance degradation factor of the hydraulic dynamometer; providing early warnings of abnormal conditions; and intelligently adjusting the test plan. The main steps include:

[0061] S01, several sensors are arranged on a turbo hydraulic dynamometer to monitor performance parameters of the hydraulic dynamometer in real time. The sensors include pressure sensors, temperature sensors, flow sensors, speed sensors, and vibration sensors. The sensor outputs are all 0-10V voltage signals. The performance parameters include rotational speed, torque, water inlet pressure, lubricating oil supply pressure, water inlet valve opening, water outlet valve opening, front bearing lubricating oil flow, rear bearing lubricating oil flow, water inlet temperature, water outlet temperature, front bearing temperature, rear bearing temperature, front bearing vibration parameters, and rear bearing vibration parameters.

[0062] S02, a hydraulic dynamometer performance parameter database is established, the sensor monitoring data is stored in the database, the 0-10V voltage signal of the sensor is converted into engineering value data corresponding to the performance parameters according to a preset conversion relationship, such as the conversion relationship of the temperature sensor, the engineering value data is screened and processed to eliminate invalid and interference data, and is stored in the database. The data storage and flow direction are as shown in Figure 4

[0063] The engineering value data screening and processing method is as follows:

[0064] Data with a rotational speed of 1000 rpm or less is eliminated as invalid data; data with a water inlet valve opening and a water outlet valve opening of less than 1 is eliminated as interference data; data with front and rear bearing temperatures higher than 70℃ can be eliminated as abnormal data in the test link to verify the model accuracy. The water outlet valve opening here refers to the water outlet valve opening, with a unit of %.

[0065] S03, according to the input requirements of the hydraulic dynamometer performance prediction model, test data is extracted according to the storage frequency of the test data in step S02.

[0066] S04, the test data extracted in step S03 is substituted into the hydraulic dynamometer performance prediction model to calculate the performance parameters of the core equipment of the hydraulic dynamometer.

[0067] S05, the relative error between the performance parameter prediction value in step S04 and the performance parameter acquisition value acquired by the corresponding sensor is calculated, and the deviation degree of the performance parameter prediction value relative to the test value of the sensor is observed by drawing a graph, which is used as a basis for determining the running state of the turbo hydraulic dynamometer. The specific deviation degree is determined by the model accuracy.

[0068] S06, according to the performance parameter acquisition value, the performance parameter prediction value, and the prediction accuracy of the hydraulic dynamometer, the running state of the hydraulic dynamometer is determined. The prediction accuracy is the difference between the calculated prediction value (performance parameter prediction value) and the experimental test value (sensor acquisition test value). The hydraulic dynamometer running state determination method in step S06 is as follows:

[0069] ​S061, if the performance parameter collected value and the performance parameter predicted value exceed the limit value, or the prediction accuracy exceeds the first threshold value, determine whether to continue testing according to the abnormal situation. If continue testing, adjust the test plan to change the test environment, and then enter step S01 to continue testing, eliminate the abnormal data at the current time, and store the abnormal data in the abnormal database table in the database. The abnormal determination result is displayed on the interface for the tester to refer to, and the running state of the hydraulic dynamometer equipment is judged. If not continue testing, end the test, and update and correct the performance prediction model according to all the health data in this test process.

[0070] S062, if the performance parameter collected value and the performance parameter predicted value do not exceed the limit value, and the prediction accuracy does not exceed the first threshold value, determine whether to continue testing according to the predetermined test plan. If continue testing, store the performance parameter predicted value and the prediction accuracy at the current time in the prediction tracking result table in the database, and enter step S01 to continue data collection and performance tracking prediction at the next time. If the current time is the final time of the test, terminate the test, and update and correct the performance prediction model according to all the health data in this test process. Here, the limit value can be divided into ultra-low limit, low limit, high limit and ultra-high limit, and the four values corresponding to the drainage temperature are 1, 5, 55 and 60 respectively. The first threshold value can be set by the user according to the actual situation.

[0071] The initial performance prediction model is obtained by training the turbine hydraulic dynamometer equipment digital twin engineering model through the initial data of the system test, which extracts the initial performance and health status of the turbine-hydraulic dynamometer system. Then, as the use time increases, due to the mutual wear and cold and hot action between system components, the system performance declines. At this time, if the initial prediction model is continued to be used, it is obviously not in line with the actual situation, and the prediction accuracy will also decrease. Therefore, at the end of each test period, the performance state of the core components of the test system needs to be evaluated, the performance degradation factor needs to be calculated, and the prediction model needs to be updated through transfer learning. Therefore, the health management method further includes step S07, performance degradation detection of the hydraulic dynamometer equipment in the performance prediction model, component performance degradation detection and degradation feature transfer learning.

[0072] Further, the step S07 includes:

[0073] S071, using the test period data at the current time, extracting the current performance state feature according to the modeling method of the hydraulic dynamometer equipment performance digital twin model, and training the current test period performance model M'. M' represents the current performance state of the hydraulic dynamometer equipment.

[0074] S072, Digitizing test on M' on original data: input state of original state is substituted into M', and performance feedback result of M' on original state is obtained. Because M' has experienced performance degradation, its feedback result will deviate from performance capability of original state, and this part of deviation reflects the level of performance degradation. The original state refers to the initial state of the dynamometer system, and the performance state of the dynamometer system is optimal at this state. With the accumulation of use time, the performance of the dynamometer system gradually degrades.

[0075] S073, Calculate component degradation factor: assume that the mean value of component core performance parameter P of the test system in original state is P0 The prediction result of M' on original state P is The degradation factor of parameter P is the difference between the two The core performance parameters in this step mainly refer to the exhaust temperature, front and rear bearing temperatures and vibration, and their degradation respectively represents the performance degradation of the water system, front and rear bearing systems.

[0076] S074, Calculate degradation factor difference, and the calculation method is the difference between the current test cycle and the previous performance parameter of each component p , that is, β α .

[0077] S075, Determine whether migration training and learning of component network layer is needed: if the difference β α is within the second threshold range, it is considered that the performance of each component has not degraded, and the current performance prediction model can continue to be used; if the difference β α exceeds the second threshold, the component exceeding the second threshold is defined as a performance degradation component, and only the network layer of the performance degradation component is migrated and trained in the subsequent performance prediction model updating process. Specifically, the weight values of the network layers of the non-degradation components in the performance prediction model are locked, and only the weight values of the network layers of the degradation components are trained and updated. The second threshold can be set by the user according to the actual situation, and in particular, the second threshold value range is preferably within 1%.

[0078] Further, the component system can also be targeted for maintenance through the change of the component degradation factor, thereby saving operation and maintenance costs. In particular, the current test cycle performance model M', the degradation factor data, and the migration learning data are saved in the database for the tester to systematically analyze the test and operation process.

[0079] In order to facilitate understanding, as Figures 5-8As shown, the application provides a digital twin performance prediction model for turbine component testing, of course, other performance prediction models can also be used. The model architecture body includes a physical information network layer S01, a physical information coupling network layer S02, and a mapping network layer S03 connected in sequence. The output end of the physical information network layer S01 is connected with the physical information coupling network layer S02, and the output end of the physical information coupling network layer S02 is connected with the mapping network layer S03.

[0080] The physical information network layer S01 includes a time information network layer S11, an inlet pipeline network layer S12, a turbine network layer S13, a front bearing network layer S14, a water system network layer S15, a rear bearing network layer S16, and an axle power system network layer S17 according to the turbine testing system.

[0081] The physical information coupling network layer S02 adopts a time sequence network structure to sequentially connect the physical information network layer S01 through time and space, extract the physical correlation and physical characteristics between the turbine hydraulic dynamometer testing parameters, such as the correlation characteristics between the rotational speed, the drainage temperature, the drainage valve, and the shaft power, and store them in the neural network. It should be noted that the input of the physical information coupling network layer S02 is the output of each sub-network of the physical information network layer S01, and the number of input network nodes of the physical information coupling network layer S02 is the number of sub-networks of the physical information network layer S01.

[0082] The mapping network layer S03 maps the information of the physical information coupling network layer S02 to the final target output value. The mapping condition is the full connection layer network mapping relationship of the neural network. The number of input nodes of the mapping network layer S03 is the number of output nodes of the physical information coupling network layer S02, and the specific value can be different according to the actual learning data amount. Preferably, the number of output nodes of the physical information coupling network layer S02 is greater than the number of input nodes, and the value is a power of 2.

[0083] Further, the arrangement of the physical information network layer S01 is sequentially set according to the working arrangement of the turbine hydraulic dynamometer system, and is sequentially set as the time information network layer S11, the inlet pipeline network layer S12, the turbine network layer S13, the front bearing network layer S14, the water system network layer S15, the rear bearing network layer S16, and the axle power system network layer S17.

[0084] Further, the time information network layer S11 is composed of time item function, mainly processing current time information and time change information, such as time or time difference value and other time-related parameters are extracted physical characteristics in the time information network layer S11. Since time plays a role in the actual operation of the test system, the time information network layer S11 is placed in the front. The time item function can be any composite parameter containing time information, such as time interval function, expressed as f = t(n) - t(n-1). Since the turbine system in the test system is a drive system, it is mainly composed of two parts of the intake pipeline and the turbine components, so the intake pipeline network layer S12 and the turbine network layer S13 are followed after the time information network layer S11.

[0085] Further, the input parameters of the intake pipeline network layer S12 are the parameters collected by the intake pipeline in the actual test system, including intake temperature and intake pressure. The input parameters of the turbine network layer S13 are the turbine core parameters collected by the sensors deployed in the turbine components, including inlet and outlet temperature, inlet and outlet pressure, turbine speed, turbine shaft oil temperature, and turbine shaft oil pressure.

[0086] Specifically, the following several front bearing network layers S14, water system network layers S15, rear bearing network layers S16, and shaft power system network layers S17 correspond to several main component systems of the turbine hydraulic dynamometer equipment. The frontmost is the front bearing network layer S14, which corresponds to the front bearing part of the turbine hydraulic dynamometer equipment, and the core input parameter is the front bearing parameter of the front end of the hydraulic dynamometer equipment, including bearing temperature, bearing oil temperature, and oil pressure. As shown in the structure principle diagram of the dynamometer equipment, Figure 1 The water system network layer S15 is arranged behind, and the input parameters include water inflow and outflow, inlet and outlet valve opening, and water temperature. The rear bearing network layer S16 corresponds to the rear bearing part of the rear end of the hydraulic dynamometer equipment, and the input parameter is the rear bearing parameter of the rear end of the hydraulic dynamometer equipment, including bearing temperature, bearing oil temperature, and oil pressure. Since the front bearing, water system, and rear bearing are driven by the dynamometer shaft, the dynamometer speed is simultaneously used as the input of the front bearing network layer S14, water system network layer S15, and rear bearing network layer S16. Since the final output of the entire test system is shaft power, the last network layer is set as the shaft power network layer S17, and the input parameters of the shaft power system network layer S17 include shaft power and shaft torque. The arrangement of the physical information network layer S01 is set according to the working arrangement of the turbine-hydraulic dynamometer system, so that the model architecture perfectly matches the physical test mechanism and structure layout.

[0087] As shown in Figure 8As shown, in order to associate each sub-network layer of the physical information network layer S01, the physical information coupling network layer S02 takes the component parameters in the axial direction as the digital medium, fuses the physical knowledge in the test system, arranges the component characteristic parameters in order by combining the network results with timing, forms the internal characteristic fusion relationship of the physical information input layer, and considers the bidirectional timing network structure in the physical information fusion relationship. It should be noted that the timing structure here can be any of the existing timing networks, including GRU, LSTM, ARIMA, and BP timing prediction model, to strengthen the interaction between component characteristic parameters and eliminate the relationship caused by the order of different measurement parameters representing the same component.

[0088] To illustrate the above bidirectional timing relationship, in terms of sequential priority, the turbine network layer will be directly affected by the positive influence of the intake pipeline network layer S12, but at the same time the turbine network layer S13 will also be inversely transmitted to the intake pipeline network layer S12, although the reverse influence is weaker than the positive influence, but this bidirectional influence cannot be removed. From the actual physical process, for example, from front to back, the pressure and temperature of the intake will change the work size of the turbine, thereby affecting the speed, and then affecting the shaft power output of the hydraulic system, while the change of the turbine component state will also cause the change of the intake pipeline airflow to a certain extent, and at the same time the shaft power value will also be used as a feedback control quantity to control the change of the intake parameter, finally making the shaft power reach the rated working state.

[0089] Further, the target output value can be any required monitoring parameter, and here the core parameter can be taken as the monitoring value, for example, taking torque or power as the core output performance monitoring value to monitor the output of the overall test performance; taking the front and rear bearing temperatures as the bearing health state monitoring value; taking the drainage flow and drainage temperature as the dynamometer heat transfer monitoring parameter.

[0090] Further, as Figure 6As shown, the physical information network layer S01 also sets up a turbine system coupling network layer S101 and a dynamometer system coupling network layer S103. The input ends of the intake pipeline network layer S12 and the turbine network layer S13 are connected with the turbine system coupling network layer S101, and the intake pipeline network layer S12 and the turbine network layer S13 perform physical sign extraction of the entire turbine system in the turbine system coupling network layer S101. The output ends of the front bearing network layer S14, the water system network layer S15 and the rear bearing network layer S16 are connected with the dynamometer system coupling network layer S103, and the front bearing network layer S14, the water system network layer S15 and the rear bearing network layer S16 perform feature extraction of the entire turbine hydraulic dynamometer system in the dynamometer system coupling network layer S103. The output ends of the turbine system coupling network layer S101 and the dynamometer system coupling network layer S103 are connected with the physical information coupling network layer S02. The turbine system coupling network layer S101 and the dynamometer system coupling network layer S103 extract features in a similar manner as the physical information coupling network layer S02.

[0091] Further, as shown in Figure 7 , the physical information network layer S01 also sets up an axle connection synchronization network layer S102, which is arranged between the turbine system coupling network layer S101 and the dynamometer system coupling network layer S103, and is used to store deviation information between the turbine system and the dynamometer system. The input information of the axle connection synchronization network layer S102 is the deviation value of the turbine shaft center and the dynamometer equipment shaft center. During the use of the dynamometer system, the shaft centers of the turbine shaft and the dynamometer equipment shaft will deviate, and the axle connection synchronization network layer S102 is mainly arranged to extract deviation features and correct the deviation.

[0092] As shown in Figure 10 , 11 , the application also discloses a digital twin model for turbine component testing based on test cumulative effect, which comprises the digital twin model for turbine component testing, a test cumulative effect network layer S104 and a coupling network layer S202. The coupling network layer S202 couples the test cumulative effect network layer S104 and the physical information coupling network layer S02 and extracts features. The extracted features are the running physical correlation features between components and subsystems in time and space, that is, the mutual influence relationship between components in space and time. The output ends of the test cumulative effect network layer S104 and the physical information coupling network layer S02 are connected with the coupling network layer S202.

[0093] Further, the test cumulative effect network layer S104 is composed of n physical information network layers of the aforementioned time points, and the system performance degradation information is transmitted from the upstream time (t_0 time point) to the downstream time (t_1 time point) by the time sequence network, and the obtained time cumulative information is taken as the output of the test cumulative effect network layer S104.

[0094] Specifically, the test cumulative effect network layer S104 obtains the time cumulative information and degradation information of the system through n physical information layers of the aforementioned time points. For example, the calculation at the n+k time point carries the cumulative information at the k-1 to n+k-1 time points, the calculation at the n+k-1 time point carries the cumulative information at the k-2 to n+k-2 time points, and so on, and the calculation at the n+k time point contains the cumulative information from the beginning of the system operation to the present.

[0095] In particular, the value of n is arbitrary, but it should not be too large, otherwise it is easy to cause cumulative information overload. At the same time, due to the irreversibility of time, the test cumulative effect network layer S104 preferably has a one-way time sequence structure. In addition, the input of the physical information network layer S01 from the t_0 time point to the t_1 time point preferably uses parameters that are strongly related to the target value, such as the target value itself.

[0096] As shown in Figure 11 The present application discloses a digital twin health management platform for turbine component testing, which comprises a collection module 6, a database 7, a data processing module 8, a performance prediction module 9, an analysis module 10, the collection module 6 is in communication connection with the data processing module 8, the performance prediction module 9 is in communication connection with the analysis module 10, and the data processing module 8, the performance prediction module 9 and the analysis module 10 are in bidirectional communication connection with the database 7. The collection module 6 is in communication connection with the database 7. The database 7 comprises a collection parameter data table, a calculation parameter data table, a prediction result tracking table, a health data table and an abnormal data table.

[0097] The collection module 6 collects the performance parameters of the hydraulic dynamometer equipment and transmits them to the data processing module 8. The collection module 6 comprises a pressure sensor, a temperature sensor, a flow sensor, a speed sensor and a vibration sensor. The performance parameters include rotational speed, torque, water inlet pressure, lubricating oil supply pressure, water inlet valve opening, water outlet valve opening, front bearing lubricating oil flow, rear bearing lubricating oil flow, water inlet temperature, water outlet temperature, front bearing temperature, rear bearing temperature, front bearing vibration parameters and rear bearing vibration parameters.

[0098] The data processing module 8 performs screening processing on the performance parameter collection values collected by the collection module 6, and stores them in the abnormal data table and the health data table in the database according to whether the data is abnormal.

[0099] The performance prediction module 9 extracts test data from the database 7 and calculates the performance parameters of the predicted hydraulic dynamometer. It should be noted that the performance prediction module 9 can not be integrated with other modules into a platform, but can be separately made into a performance prediction digital platform, and interact with other modules. It can also be designed on the same management platform as other modules.

[0100] The analysis module 10 extracts calculation data from the database 7, calculates the relative error between the predicted value of the performance parameter and the collected value of the performance parameter collected by the sensor, and judges the running state of the hydraulic dynamometer.

[0101] An electronic device, comprising one or more processors; a storage device for storing one or more programs; when the one or more programs are executed by the one or more processors, the one or more processors implement a digital twin health management method for turbine component testing.

[0102] A computer readable medium, the readable medium storing a computer program, the computer program being executed by a processor to implement a digital twin health management method for turbine component testing.

[0103] Embodiment one

[0104] S01: The technical personnel arrange pressure sensors, temperature sensors, flow sensors, speed sensors and vibration sensors on the turbine hydraulic dynamometer entity to monitor the key performance parameters of the turbine hydraulic dynamometer in real time, such as speed, torque, water inlet pressure, oil supply pressure, water inlet valve opening, water outlet valve opening, front bearing oil flow, rear bearing oil flow, water inlet temperature, water outlet temperature, front bearing temperature, rear bearing temperature, front bearing vibration parameter and rear bearing vibration parameter. The turbine hydraulic dynamometer health management platform and the performance prediction digital platform will monitor the measurement values of all sensors in real time, and will issue an alarm on the platform UI interface when any sensor output data is abnormal, reminding the technical personnel to check whether the sensor hardware is working normally. Abnormal data conditions, such as sudden rise or fall of values, too large difference between the value at a certain time and the values before and after that time, or 0 or inf output value of the sensor, etc.

[0105] S02: The performance prediction digital platform and health management platform converts the received 0-10V voltage signal into the corresponding actual engineering value according to the preset conversion relationship, and stores it in the acquisition parameter data table in the hydraulic dynamometer system database. These data are further screened and cleaned to exclude abnormal data that may occur during data conversion. For example: for data that suddenly rises, falls and returns to normal value only at a certain time, the average value of the previous and subsequent time is replaced; for data with a value of inf, the limit value is replaced; for data less than zero, 0 is uniformly replaced. The data processed are saved in the calculation parameter data table.

[0106] S03-S05: In this step, the modeling method of the performance digital twin model of the hydraulic dynamometer is built, specifically, the test data at the beginning of the test system operation is used as the input and output data source of the performance digital twin model, the initial state performance prediction model is trained, and then the prediction model is updated after each test period, so the prediction model in this step is the updated prediction model after the last test period. The prediction model extracts the input parameters of the model from the data table according to the saving frequency or preset frequency of the data in the calculation parameter data table, calculates and predicts the performance prediction parameters of the key components, and calculates the prediction error according to the following formula:

[0107]

[0108] In the formula, X p is the predicted value of parameter X, and X t is the test value of parameter X.

[0109] In particular, it needs to be emphasized again that the performance digital twin model of the turbine hydraulic dynamometer equipment is constructed based on data and the physical operation law of the turbine-hydraulic dynamometer system, and has the advantages of high precision, strong physical interpretation and small amount of required data.

[0110] S06: Specifically, the abnormal state in step S06 in the present application includes but is not limited to: data within 1000 rpm can be excluded as invalid data. The valve position fluctuation exceeding ±4% can be considered as a poor working state or near the boundary. The data with large bearing temperature change rate can be considered as a fault state. The drainage temperature higher than 60℃ can be considered as an abnormal working state.

[0111] Through actual abnormal state data test, when the running state is abnormal, such as the rear bearing temperature over-temperature, the prediction error will obviously increase, so when the actual prediction error (a ± s) obviously increases and exceeds the threshold value d, it can be considered that the current hydraulic dynamometer is in an unstable working state. It should be noted that the average error of the prediction model test set, that is, the average error reflected on the artificially selected test data set after the completion of the prediction model training, the actual error calculated in the running process of the turbine hydraulic dynamometer health management software is the instantaneous error, which will fluctuate up and down on a, and the fluctuation range is ± s.

[0112] The digital platform extracts the turbine hydraulic dynamometer operating parameters, key performance parameters and their prediction values and relative errors from the database, and plots and displays them on the UI interface, as shown in FIG. 8. Figure 5 According to the preset limit value, the turbine hydraulic dynamometer health management platform and the performance prediction digital platform can monitor the running state of the turbine hydraulic dynamometer in real time, and alarm when the limit is exceeded, and the technician can adjust the test plan in real time according to the prediction result.

[0113] The turbine hydraulic dynamometer health management platform and the performance prediction digital platform proposed in the present application need the technician to preset the following parameter indicators in advance when actually deployed: the conversion relationship between the voltage signal corresponding to the sensor and the actual physical value; the limit value of the operating parameter that needs to be monitored in real time, including the lower limit, the ultra-low limit, the upper limit and the ultra-high limit. The operating parameter that needs to be predicted by the technician.

[0114] Before using the hydraulic dynamometer to conduct experiments on turbine components, the turbine hydraulic dynamometer health management and performance prediction digital platform needs to be opened in advance, and the test is conducted after clicking start test. The variables that can be displayed or switched by the technician include: the real-time change curve of the monitoring parameters displayed by the turbine hydraulic dynamometer health management and performance prediction digital platform, the parameters and the range can be switched in real time; the prediction frequency of the operating parameter that needs to be predicted by the technician, which is 2HZ by default; the display range of the prediction parameter, including self-adaptation and custom upper and lower limits; the key information that needs to be recorded in this experiment is input in text form on the UI interface and saved in the hydraulic dynamometer database together with the data; past test data records, the past hydraulic dynamometer test data stored in the database can be renamed, deleted and other operations on the UI interface.

[0115] The parameter selection of the platform described in the present application is based on the results of artificial experience and correlation analysis, and the key performance parameters of the key components in the turbine-dynamometer system are selected as the monitoring values. In particular, the drain temperature, torque, front and rear bearing temperature and vibration are preferably selected as the monitoring parameters.

[0116] Particularly, as described in Embodiment One, the data interaction of the platform is realized through a database, as shown in Figure 5 That is, a hydraulic dynamometer database is established in the platform workstation, and a collection parameter data table, a calculation parameter data table, a prediction tracking result table, and an abnormal data table are established, so that the collection parameters and prediction results during the entire test operation process are classified and saved.

[0117] The above-described embodiments are only used to describe the preferred modes of the present application, and do not limit the scope of the present application. Without departing from the design spirit of the present application, various modifications, variations, modifications, and replacements of the technical solutions of the present application made by those skilled in the art shall fall within the protection scope determined by the claims of the present application.

Claims

1. A digital twin health management method for turbine component testing, characterized by: The method comprises the steps of: S01, arranging a plurality of sensors on a turbine hydraulic dynamometer device to monitor the performance parameters of the hydraulic dynamometer device in real time; S02, establishing a hydraulic dynamometer device performance parameter database, storing the sensor monitoring data in the database, converting the voltage signal of the sensor into the engineering value data corresponding to the performance parameter according to a preset conversion relationship, and storing the engineering value data in the database after screening processing; S03, extracting test data according to the storage frequency of the sensor monitoring data in step S02 according to the input requirements of the hydraulic dynamometer device performance prediction model; S04, inputting the test data extracted in step S03 into the hydraulic dynamometer device performance prediction model to calculate the predicted performance parameters of the hydraulic dynamometer device; S05, calculating the relative error between the predicted performance parameters in step S04 and the performance parameter acquisition values acquired by the sensor; S06, judging the running state of the hydraulic dynamometer device according to the performance parameter acquisition values, the performance parameter prediction values and the prediction accuracy of the hydraulic dynamometer device; S07, performing transfer learning on the hydraulic equipment performance degradation detection, component performance degradation detection and degradation characteristics in the performance prediction model; The step S07 comprises, S071, extracting the current performance state characteristics according to the hydraulic dynamometer device performance digital twin model using the current test period data, and training the current test period performance model M'; S072, performing digital test on M' on the original data: substituting the input state of the original state into M' to obtain the performance feedback result of M' on the original state; S073、Calculate component degradation factor: assume the mean value of component core performance parameter P in steady state operation under the original state of hydraulic testing equipment , the prediction result of M' to the original state P is , then the degradation factor of parameter P is ; S074、Calculate the degradation factor difference: the difference between the performance parameter of each component in the current test cycle and the previous one difference value; S075, judging whether the component network layer needs to be transferred and trained: If the difference is within a second threshold range, it is considered that the performance of each component has not degraded, and the current performance prediction model can continue to be used. If the difference is within a second threshold range, it is considered that the performance of each component has not degraded, and the current performance prediction model can continue to be used. If If the difference exceeds the second threshold, the component exceeding the second threshold is defined as a performance degradation component, and only the network layer of the performance degradation component is trained and learned in the subsequent performance prediction model updating process. The architecture of the performance prediction model comprises three layers of physical information network layer, physical information coupling network layer and mapping network layer connected in sequence, the output end of the physical information network layer is connected with the physical information coupling network layer, and the output end of the physical information coupling network layer is connected with the three layers of mapping network layer; The physical information network layer comprises time information network layer, inlet pipeline network layer, turbine network layer, front bearing network layer, water system network layer, rear bearing network layer and shaft power system network layer according to the turbine test system; The physical information coupling network layer adopts a time sequence network structure, connects the physical information network layer through time and space in sequence, extracts the physical correlation and physical characteristics between the test parameters of the turbine hydraulic dynamometer device, and stores them in the neural network; The mapping network layer maps the information of the physical information coupling network layer to the final target output value, and the mapping condition is the full connection layer network mapping relationship of the neural network.

2. The digital twin health management method for turbine component testing of claim 1, wherein: The sensors in step S01 include pressure sensors, temperature sensors, flow sensors, speed sensors and vibration sensors; and / or the performance parameters include rotational speed, torque, water inlet pressure, lubricating oil supply pressure, water inlet valve opening, water outlet valve opening, front bearing lubricating oil flow, rear bearing lubricating oil flow, water inlet temperature, water outlet temperature, front bearing temperature, rear bearing temperature, front bearing vibration parameter and rear bearing vibration parameter.

3. The digital twin health management method for turbine component testing of claim 2, wherein: The step S02 in the engineering value data screening processing method is as follows: data within a rotating speed of 1000 rpm is removed as invalid data; data with a water inlet and outlet valve position lower than 1% is removed as interference data; data with a front and rear bearing temperature higher than 70 DEG C is removed as abnormal data from the training data, and the model accuracy is verified in the test link.

4. The digital twin health management method for turbine component testing of claim 1, wherein: The step S06 in the water power dynamometer equipment running state judgment method is as follows, S061, if the performance parameter acquisition value and the performance parameter prediction value exceed the limit value, or the prediction accuracy exceeds the first threshold value, whether to continue testing is determined according to the abnormal situation; If continue testing, change the test environment and enter step S01 to continue testing, remove the abnormal data at the current time and store the abnormal data to the abnormal database table in the database; If not continue testing, end the test, and update and correct the performance prediction model according to all the health data in the test process; S062, if the performance parameter acquisition value and the performance parameter prediction value do not exceed the limit value, and the prediction accuracy does not exceed the first threshold value, whether to continue testing is determined according to the predetermined test plan; If continue testing, the performance parameter prediction value and the prediction accuracy at the current time are stored to the prediction tracking result table in the database, and step S01 is entered to continue data acquisition and performance tracking prediction at the next time; If the current time is the final time of the test, terminate the test, and update and correct the performance prediction model according to all the health data in the test process.

5. A management platform for a digital twin health management method for turbine component testing according to any one of claims 1-4, characterized in that: The system comprises a collection module, a database, a data processing module, a performance prediction module, and an analysis module, wherein the collection module is in communication connection with the data processing module, the performance prediction module is in communication connection with the analysis module, and the data processing module, the performance prediction module, and the analysis module are connected with the database; The collection module collects performance parameters of the water power dynamometer equipment and transmits the performance parameters to the data processing module; The data processing module performs screening processing on the performance parameter acquisition value, and stores the performance parameter acquisition value to the abnormal data table and the health data table in the database according to whether the performance parameter acquisition value is abnormal; The performance prediction module extracts test data from the database, and calculates the performance parameters of the water power dynamometer equipment; The analysis module calculates the relative error between the performance parameter prediction value and the performance parameter acquisition value collected by the sensor, and judges the running state of the water power dynamometer equipment.

6. The digital twin health management platform for turbine component testing of claim 5, wherein: The collection module comprises a pressure sensor, a temperature sensor, a flow sensor, a speed sensor, and a vibration sensor. The performance parameters comprise rotating speed, torque, water inlet pressure, lubricating oil supply pressure, water inlet valve opening, water outlet valve opening, front bearing lubricating oil flow, rear bearing lubricating oil flow, water inlet temperature, water outlet temperature, front bearing temperature, rear bearing temperature, front bearing vibration parameter, and rear bearing vibration parameter.

7. An electronic device, comprising: The system comprises: one or more processors; a storage device for storing one or more programs; when the one or more programs are executed by the one or more processors, the one or more processors implement the digital twin health management method for turbine component testing according to any one of claims 1-4.

8. A computer readable medium storing a computer program, characterized in that: The computer program, which is executed by a processor, implements the method for digital twin health management for turbine component testing as claimed in any of claims 1-4.

Images