CN105134456A - Water turbine fault prognosis method based on on-line monitoring - Google Patents

Abstract

A hydraulic turbine fault pre-diagnosis method based on online monitoring, through the collection of hydraulic turbine vibration data samples, real-time database reading and writing, data preprocessing, and formulation of reasoning rules, the online monitoring of hydraulic turbine vibration status and real-time fault pre-diagnosis are realized. It includes the collection and display of real-time data, the realization of fault warning and vibration trend display, the recording and display of diagnosis results, fault analysis, and treatment suggestions, and the storage of diagnosis reports.

Description

Hydro Turbine Fault Prediction Method Based on Online Monitoring

technical field

The invention relates to the field of industrial control, in particular to a method for diagnosing vibration faults of hydraulic generator sets.

Background technique

For a long time, in hydroelectric generating units, mechanical vibration has been a core factor that threatens the safe production and operation of hydroelectric generating units. The vibration of hydropower units is often caused by the joint action of mechanical, electrical and hydraulic factors, and the vibration mechanism is relatively complex. The vibration and swing of the unit will also be caused by design, installation, operation and other reasons, which cannot be completely avoided and eliminated. In short, general vibration will not cause harm to the unit, but if it exceeds the allowable value seriously, especially for a long time, it will have a serious impact on the unit.

At present, most diagnostic techniques are based on expert systems, and it is impossible to make a diagnosis and issue an early warning for real-time data by analyzing and diagnosing historical data. This application proposes a real-time diagnostic method.

Contents of the invention

The invention proposes a hydraulic turbine fault pre-diagnosis method based on online monitoring, which belongs to the real-time diagnosis method of hydraulic turbines, and solves the problem that the traditional diagnostic method cannot perform online diagnosis.

The present invention specifically adopts the following technical solutions.

A method for pre-diagnosing hydraulic turbine faults based on online monitoring, said method comprising the following steps:

Step 1: Collect real-time data of the turbine in real time through the online monitoring device, including the vibration value and amplitude of the upper frame of the turbine in the X and Y directions, the swing value and amplitude of the upper guide in the X and Y directions, and the water guide in the X and Y directions Swing value and amplitude, speed, unit load, unit load rejection signal, unit vibration status signal;

Step 2: Perform the following preprocessing on the data collected in real time in step 1:

Save the collected data values in the data list according to the station number of the hydropower station, the turbine unit number and the id number in the real-time database;

The collected data values are compared according to value>va, and the diagnostic condition identification value ans is obtained according to the comparison result, and the diagnostic condition identification value ans is saved in the data list, where value is the collected real-time data value, and va is the value corresponding to a certain The preset alarm value of similar real-time data, ans is the diagnostic condition identification value, the value is 0 or 1, 0 means value<va, 1 means value>va;

Step 3: First, build a knowledge base based on the experience of operating personnel and experts. The knowledge base includes the specific content of diagnostic conditions, diagnostic results, fault analysis, handling suggestions, and rule numbers. Use the specific description corresponding to the diagnostic condition identification value ans in step 2 as The content of the fault analysis, and the processing suggestions are the processing measures for the fault. Each piece of knowledge includes the above content; The form is as follows:

When AANDBANDCAND..., the fault diagnosis result R is obtained; among them, A, B, C... are based on the comparison of various real-time data values collected in step 1 with the preset alarm values of corresponding real-time data in step 2 The obtained diagnostic condition identification value, that is, the diagnostic condition identification value ans in step 2, wherein, AND indicates that the logical relationship between each diagnostic condition identification value is "and", and R indicates the final fault diagnosis result, that is, when all kinds of real-time The diagnostic condition identification values obtained from the data are not all 1, indicating that the diagnosis result is no fault, and when the diagnostic condition identification values obtained from various real-time data are all 1, it indicates that the diagnosis result is fault.

For example: the real-time data of the vibration of the upper frame in the X direction in step 1, after the judgment in step 2, the corresponding diagnostic condition identification value A is obtained. When A is 1, it means exceeding the standard, and when A is 0, it means normal. Take the description of the identification value of the corresponding diagnostic condition as the content of the fault analysis, such as: the X-direction vibration value of the upper frame exceeds the standard; and provide guidance as a processing suggestion, such as: please check whether the bolts of the upper frame are loose.

Step 4: After the data processing in step 2 and the fault diagnosis in step 3, after obtaining the diagnosis results, fault analysis and treatment suggestions, display the diagnosis results, fault analysis and treatment suggestions through the man-machine interface, and query the collected turbine data The real-time and historical curves, the diagnosis results, fault analysis and treatment suggestions are stored as a diagnosis report, and the diagnosis time is recorded; finally, the logical relationship of the inference rule is displayed on the man-machine interface at the same time.

As mentioned above, it is the basic process of the diagnosis and pre-diagnosis method. The advantages of this method are as follows:

1. Online monitoring, real-time diagnosis: real-time diagnosis by collecting real-time data, and drawing real-time data curves and historical data curves;

2. The knowledge base is easy to maintain: the knowledge base is saved in XML format, and knowledge can be added, modified and deleted at any time;

3. Convenient query of diagnosis results: the diagnosis method finally outputs a diagnosis report, which is saved in text form, and the content includes diagnosis criteria, diagnosis results, fault analysis, treatment suggestions and diagnosis time (accurate to seconds);

4. The man-machine interface is rich in content: the man-machine interface not only displays the diagnosis results, conclusion analysis, treatment suggestions, and diagnosis logic, but also can monitor data, query real-time curves and historical curves.

Description of drawings

Fig. 1 is a schematic diagram of data flow and fault pre-diagnosis flow of the present invention;

Detailed ways

The technical solution of the present invention will be further described in detail below in conjunction with the accompanying drawings. Figure 1 shows the diagnostic data collection and diagnostic flow chart.

Step 1. According to the needs of on-site application services, the vibration sample data of each monitoring channel of the hydro-generator set can be collected through the vibration monitoring device, and then the sample data generated by the vibration device can be obtained through a dedicated device and a customized driver, so as to realize various The sample data of the vibration channel is connected to the real-time database, mainly: the vibration value and amplitude of the upper frame of the turbine in the X and Y directions, the swing value and amplitude of the upper guide in the X and Y directions, and the swing of the water guide in the X and Y directions Value and amplitude; and collect other relevant data: speed, unit load, unit load rejection signal, unit vibration status signal.

Step 2, data preprocessing. First, the fault diagnosis module will obtain the sample data collected by the vibration device from the real-time database. The method of obtaining is: to establish a model file corresponding to the real-time data. The format of the file is XML, and the content includes: Name and unit number, point number id, point name, description, data value and data reading times of the collected data points, the name of the hydropower station is represented by the "stationname" field, and the corresponding "id" is the plant station number; it is represented by "hydroturbinename" The name of the water turbine unit, corresponding to "id" is the unit number; "PropItemname" indicates the name of the sample data point, corresponding to "id" indicates that the point stores the id number in the real-time database, and "cValue" indicates the alarm value or limit of the data point Value, "flag" is the identification of the number of times the data is read. Through the real-time library access interface provided by the monitoring platform, data is collected according to the plant station number, unit number, and point number, and the data is saved in the data list according to the content of the model file, including the point name, description, and data value of the point.

After acquiring real-time data, the data is first sent to the man-machine interface for display through the interface provided by the diagnostic module (the displayed content includes roll call, description, real-time value, and alarm value), and then the data is preprocessed, and the collected data are matched according to value>va The data described in step 1 is calculated, and the diagnostic condition identification value ans is defined according to the calculation result. If value<va, the value of ans is 0; if value>va, the value of ans is 1, and the value of ans is saved in the data list, where value is the collected real-time data value, and va is the alarm value, which is used by the inference engine.

Step 3. Firstly, according to the experience of operating personnel and experts, a knowledge base is established. The knowledge base includes the specific content of diagnostic conditions, diagnostic results, fault analysis, handling suggestions and rule numbers. The specific description corresponding to the diagnostic condition identification value ans in step 2 is taken as The content of the failure analysis, the handling suggestion is the handling measure for the failure, and each piece of knowledge includes the above content;

Then, use the diagnostic condition identification value ans obtained in step 2 to perform fault diagnosis, the specific process is as follows:

1. First define a set of intermediate variables identified by preconditions, use the list obtained after data preprocessing to read the value of the precondition identifier "ans" and assign it to the intermediate variables;

2. Use precondition intermediate variables for inference combination to form inference rules, define and record the rule number, and then use the rule number to obtain the content of the knowledge corresponding to the rule number in the knowledge base through the interface function written in Java code. If the content of a piece of knowledge can be completely obtained, a diagnosis is completed. If the content in the knowledge base cannot be obtained through the rule number, it will prompt for knowledge supplementation, and return to data acquisition for the next diagnosis;

3. After completing a diagnosis, read the system time to determine the time when the failure occurred, and record and store the time;

Step 4. After the above three steps, the diagnosis results, fault analysis, and treatment suggestions will be obtained, and the diagnosis conclusions, fault analysis, and treatment suggestions will be sent to the man-machine interface for display through the human-computer interaction interface function provided by the inference engine, and at the same time Display the logical relationship of this rule. The logical relationship is to read the corresponding diagnostic conditions and diagnostic results in the knowledge base through the interface function, and combine and display the diagnostic conditions according to the logical "AND" relationship of the inference rules, and "data acquisition The data sent to the human-machine interface in the "stage together form the final human-machine interface, and the real-time curve and historical curve of the diagnostic data can be queried; then, the diagnostic results, fault analysis, treatment suggestions, diagnostic condition description and The recorded fault occurrence time is saved in the specified text file as a diagnosis report.

Claims (1)

1. A hydraulic turbine fault pre-diagnosis method based on on-line monitoring, said method comprising the following steps: Step 1: Collect real-time data of the turbine in real time through the online monitoring device, including the vibration value and amplitude of the upper frame of the turbine in the X and Y directions, the swing value and amplitude of the upper guide in the X and Y directions, and the water guide in the X and Y directions Swing value and amplitude, speed, unit load, unit load rejection signal, unit vibration status signal; Step 2: Perform the following preprocessing on the real-time data of the turbine collected in step 1: Save the collected data values in the data list according to the station number of the hydropower station, the turbine unit number and the id number in the real-time database; The collected data values are compared according to value>va, and the diagnostic condition identification value ans is obtained according to the comparison result, and the diagnostic condition identification value ans is saved in the data list, where value is the collected real-time data value, and va is the value corresponding to a certain The preset alarm value of similar real-time data, ans is the diagnostic condition identification value, the value is 0 or 1, 0 means value<va, 1 means value>va; Step 3: Based on the experience of operating personnel and experts, establish a knowledge base, which includes the specific content of diagnostic conditions, diagnostic results, fault analysis, handling suggestions and rule numbers, and use the specific description corresponding to the diagnostic condition identification value ans in step 2 as the fault The content of the analysis and the processing suggestion are the processing measures for the fault; then, according to the diagnostic condition identification value obtained in step 2, the logical reasoning method based on inference rules is used to diagnose the fault, and its basic form is as follows: When AANDBANDCAND..., the fault diagnosis result R is obtained; among them, A, B, C... are based on the comparison of various real-time data values collected in step 1 with the preset alarm values of corresponding real-time data in step 2 The obtained diagnostic condition identification value, that is, the diagnostic condition identification value ans in step 2, wherein, AND indicates that the logical relationship between each diagnostic condition identification value is "and", and R indicates the final fault diagnosis result, that is, when all kinds of real-time The diagnostic condition identification values obtained from the data are not all 1, indicating that the diagnosis result is no fault, and when the diagnostic condition identification values obtained through various real-time data are all 1, it indicates that the diagnosis result is fault; Step 4: After the data processing in step 2 and the fault diagnosis in step 3, after obtaining the diagnosis results, fault analysis and treatment suggestions, display the diagnosis results, fault analysis and treatment suggestions through the man-machine interface, and query the collected turbine data Real-time curves and historical curves, the diagnosis results, fault analysis and treatment suggestions are stored as a diagnosis report, and the diagnosis time is recorded; finally, the logical relationship of the reasoning rule is displayed on the man-machine interface at the same time.