CN105759784A - Fault diagnosis method based on data envelopment analysis - Google Patents

Abstract

The invention relates to a fault diagnosis method based on data envelope analysis, which is characterized in that it comprises the following steps: inputting multiple sets of original test data; performing time alignment processing on multiple sets of original test data to generate multiple sets of original test data on the envelope , Offline; input the test data that needs to be diagnosed; judge whether the test data is in the envelope area of the upper and lower lines of the envelope, if it is in the envelope area, it is judged as qualified, if there is data outside the envelope area, It is judged as a failure. The invention performs systematic historical data envelope analysis based on a large amount of historical data accumulated by the launch vehicle, effectively solves the modeling bottleneck of the traditional fault diagnosis method, and solves the problem of complex modeling of the traditional fault tree and heavy workload.

Description

A Fault Diagnosis Method Based on Data Envelope Analysis

technical field

The invention relates to a fault diagnosis method based on data envelopment analysis, which belongs to the field of fault diagnosis.

Background technique

During the test and launch process of the launch vehicle, it is necessary to use the fault diagnosis system to monitor its condition. The traditional fault diagnosis methods mostly adopt the diagnosis method based on the fault tree. On the basis of establishing the fault tree model of the object, when the object is actually running, the fault search method is used to search the fault tree and complete the fault diagnosis.

Commonly used fault search methods include logical reasoning (using a top-down search method, starting from the top event of the fault tree, first testing the initial intermediate event, and then testing the next level of intermediate events according to the test results of the intermediate event, until the test bottom event, complete the fault diagnosis) and the minimum cut set method (test the minimum cut set corresponding to the fault mode in the fault tree one by one, complete the fault diagnosis).

The fault tree diagnosis method has the advantages of intuitive and simple diagnosis methods, but the core and key of the fault tree diagnosis method is to construct a correct and comprehensive fault tree, and the fault tree modeling is just the bottleneck of fault diagnosis, which usually requires a very comprehensive fault tree analysis. Establishing a fault tree with a complete system requires a huge workload and consumes a lot of manpower and material resources, and once the fault tree is established incorrectly or incompletely, the fault tree diagnosis method will fail to a large extent.

Contents of the invention

The purpose of the present invention is to overcome the deficiencies of the prior art, to provide a fault diagnosis method based on data envelope analysis, which breaks through the modeling bottleneck of traditional fault diagnosis methods, and can automatically generate data envelopes through the data envelope analysis method, Perform real-time comparison of test data to complete fault parameter detection and fault location functions.

The object of the invention is achieved through the following technical solutions:

Provide a kind of fault diagnosis method based on data envelopment analysis, it is characterized in that comprising the steps:

(1) Input multiple sets of original test data;

(2) Perform time alignment processing on multiple sets of original test data, and generate the envelope upper and lower lines of multiple sets of original test data;

(3) Input the test data that needs to be diagnosed;

(4) Judging whether the test data is in the envelope area of the upper and lower lines of the envelope, if it is in the envelope area, it is judged as qualified, and if there is data outside the envelope area, it is judged as failure.

Preferably, the method for generating the upper and lower envelopes of multiple groups of original test data in the step (2) is: for any time t, get the maximum value in multiple groups of original test data to form the upper line of the envelope; take multiple groups of original The minimum value in the test data forms the lower line of the envelope;

Preferably, input multiple groups of original test data in the step (1), and set up the database instance of MySQL database; The data in the MySQL database instance is imported in the Hive tool; Realize multiple groups of original test data by Hive tool and carry out time alignment processing and Generate envelope upper and lower lines of multiple sets of original test data.

Preferably, the import task is performed by using the open source component sqoop in Hive.

Preferably, in the step (2), multiple groups of original test data are subjected to time alignment processing, and the specific method for generating the upper and lower envelopes of multiple groups of original test data is: divide the sampling time into N intervals with equal time , taking the maximum value of multiple sets of original test data in each interval, and performing straight line fitting to form the upper envelope line; taking the minimum value among multiple sets of original test data in each interval, performing straight line fitting to form the lower envelope line.

Preferably, in the step (2), multiple groups of original test data are time-aligned, and the specific method for generating the upper and lower envelopes of multiple groups of original test data is: respectively fitting multiple groups of original test data to obtain Multiple fitting curves; select a set of sampling data as a benchmark, calculate the maximum value of each point on the multiple fitting curves corresponding to each sampling time of the set of sampling data, and perform fitting to form an upper line of the envelope; calculate the set of sampling data Each sampling time corresponds to the minimum value of each point on multiple fitting curves, and fitting is performed to form the lower line of the envelope.

Preferably, the multiple sets of original test data entered in step (1) select the original test data within one year.

Preferably, if it is judged to be qualified, step (5) is also included, inputting the test data to be diagnosed into the MySQL database.

Preferably, the multiple sets of original test data are more than 5 sets.

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

(1) The fault diagnosis method based on data envelope analysis of the present invention carries out systematic historical data envelope analysis according to a large amount of historical data accumulated by the launch vehicle, effectively solves the modeling bottleneck of the traditional fault fault diagnosis method, and solves the traditional fault tree The problem of complex modeling and heavy workload; the traditional fault diagnosis method relies on expert experience and needs to ensure the integrity of the fault tree. The fault diagnosis method of the present invention is based on the historical test data of the qualified test and does not need to establish a fault tree, which reduces the Difficulty of fault diagnosis;

(2) The present invention conducts analysis based on multiple sets of historical test data, and the probability of misjudgment is low.

(3) The data alignment processing of the present invention can truly reflect the volatility of the original data, and improve the accuracy of the upper and lower lines of the envelope.

(4) The present invention adopts the Hive tool to realize multiple groups of original test data to perform time alignment processing and generate envelopes on and off the line, with fast analysis speed and strong real-time processing capability.

(5) The present invention adopts an intelligent alarm beyond the enveloping area, and realizes intelligent fault diagnosis.

Description of drawings

Fig. 1 is a schematic diagram of the overall process planning of Fig. 1 of the present invention;

Fig. 2 is a schematic diagram of data preprocessing alignment in Fig. 2 of the present invention;

Fig. 3 is a schematic diagram of fault diagnosis based on the envelope of the present invention;

Fig. 4 is a schematic diagram of sending A and B data of the present invention;

Fig. 5 is a schematic diagram of envelope analysis according to an embodiment of the present invention.

detailed description

1. Task process planning

The overall workflow is shown in Figure 1. The original data is stored in the form of a DAT file. First, the text is parsed and imported into the MySQL database, and then transferred to the Hive database. After data preprocessing and data analysis, the information in the historical data is mined and stored in the In the HBase database, the web server queries HBase and returns the required data.

2. Enter the original data

The original data is organized according to the hierarchical structure of models, dispatch times, and parameters. Each DAT file usually has 2-3 columns, ranging from thousands to hundreds of thousands of lines, and is imported into MySQL in batches through the Python program.

3. Transfer data to Hive

Hive is a data warehouse tool based on Hadoop, which can map structured data files into a database table and provide SQL query functions. In order to facilitate the work of the big data mining program, the MySQL data is imported into Hive, and the open source component sqoop is used to execute the import task in parallel.

4. Data preprocessing and envelope analysis

The data sampling time of different transmissions of the same model is not completely matched. For subsequent envelope analysis, the data needs to be processed to be strictly aligned according to time, while ensuring the authenticity of the data. Figure 2 shows the interval alignment algorithm.

The data envelopment analysis method is a method of abnormal parameter diagnosis based on historical data. By calculating the historical data of the relative time in the same moment in the historical test data of multiple rockets, the upper limit and the lower limit of the envelope are obtained. Finally, all the The upper limit time points of the envelope are connected into a curve to generate the upper limit of the envelope, all the time points of the lower limit of the envelope are connected into a curve to generate the lower limit of the envelope, and finally the envelope domain is generated. When performing fault diagnosis based on data envelope analysis, pass The measured value of the test data is compared with the envelope domain to judge whether the test parameters are abnormal. Envelope analysis selects historical parameter data similar to multiple missions, and determines the historical envelope of the parameter under the task by finding the maximum and minimum values at each moment. Using the big data platform, a large amount of historical data can be comprehensively analyzed to obtain reliable data. The parameter history envelope that reflects the real situation is convenient for timely monitoring of abnormal parameters in the future.

However, since the tasks of each round may be different, the same parameter may show different characteristics under different rounds. If the envelope is randomly selected to generate an envelope, it will cause the envelope to degenerate, and the upper and lower bounds are relatively wide. It is difficult to reflect the characteristics of specific parameters, as shown in Figure 3.

Therefore, when generating the envelope, you should choose a close occurrence, such as using data within one year, and strive to make the envelope reflect the real and common situation without degrading its own characteristics. By calculating the envelope contribution and abnormal envelope, irrelevant occurrences are eliminated.

The first alignment method is: divide the sampling time into N intervals with equal time, take the maximum value of multiple sets of original test data in each interval, and perform straight line fitting to form the upper line of the envelope; take multiple sets of each interval The minimum value in the original test data is fitted with a straight line to form the lower line of the envelope.

The second alignment method is as follows: Fit multiple sets of original test data to obtain multiple fitting curves; select a set of sampling data as a benchmark, and calculate each sampling time of the set of sampling data corresponding to each of the multiple fitting curves. The maximum value of the point is fitted to form the upper line of the envelope; the minimum value of each point on the multiple fitting curves corresponding to each sampling time of the group of sampling data is calculated, and the fitting is performed to form the lower line of the envelope.

5. Front-end display

Display the upper and lower envelopes and diagnostic information. In order to facilitate the viewing of data analysis results, the results are displayed in B/S mode. UWSGI is used as the web server, the browser's request is forwarded to the webService written in Python, the HBase database is accessed using the Happybase module, and the query results are returned to the browser via uWSGI.

The method of data envelopment analysis is realized by program, a plug-in is generated, integrated into the big data platform, and finally a fault diagnosis system based on data envelopment analysis is generated.

6. Embodiment

Through the big data platform, the flight data is entered into the big data platform, and the envelope analysis program is added to the big data platform to form a strong analysis capability. It is verified by the data of a certain model. See Figure 4. There is an abnormal situation in the known flight A. , take this issue A as the verification data, and use the other several times of data as the historical basic data to compare with the conventional analysis to verify the analysis function of the platform. The test parameter selects a certain temperature (XX) in the existing analysis report.

The envelope analysis of parameter XX is shown in Fig. 5. It can be seen from the routine analysis diagram that at around 1240s, the parameter XX is abnormal compared with B. It can be seen from the envelope analysis diagram that the A parameter of the transmission time slightly crossed the boundary around 580s, and seriously crossed the boundary around 1240s.

Through the comparative analysis of the parameter XX, it can be seen that the conventional analysis method fails to judge the fault of about 500s, and can only give a fault judgment after a serious abnormality occurs in 1240s. The envelope analysis method can judge the fault of about 500s.

The invention is an effective means to realize the real-time on-line fault detection and location of the launch vehicle test and launch, and at the same time greatly promotes the application of the fault diagnosis system in the launch vehicle industry, and has broad application prospects and huge market potential.

The above description is only the best specific implementation mode of the present invention, but the scope of protection of the present invention is not limited thereto. Any person skilled in the art can easily conceive of changes or modifications within the technical scope disclosed in the present invention. Replacement should be covered within the protection scope of the present invention.

The content that is not described in detail in the specification of the present invention belongs to the well-known technology of those skilled in the art.

Claims (9)

1. A fault diagnosis method based on data envelopment analysis, characterized in that comprising the steps: (1) Input multiple sets of original test data; (2) Perform time alignment processing on multiple sets of original test data, and generate the envelope upper and lower lines of multiple sets of original test data; (3) Input the test data that needs to be diagnosed; (4) Judging whether the test data is in the envelope area of the upper and lower lines of the envelope, if it is in the envelope area, it is judged as qualified, and if there is data outside the envelope area, it is judged as failure. 2. according to the described method of claim 1, it is characterized in that: in the described step (2), the envelope method of generating multiple groups of original test data on and off the line is: for any moment t, get the maximum in multiple groups of original test data value to form the upper line of the envelope; take the minimum value among multiple sets of original test data to form the lower line of the envelope. 3. according to the described method of claim 1, it is characterized in that: in the step (1), enter multiple groups of original test data, and set up the database example of MySQL database; The data in the MySQL database example is imported in the Hive tool; In Hive tool Realize multiple sets of original test data for time alignment processing and generate multiple sets of original test data on and off the envelope. 4. The method according to claim 3, characterized in that: utilizing the open source component sqoop in Hive to perform the import task. 5. according to the described method of claim 1, it is characterized in that: in described step (2), multiple groups of original test data are carried out time alignment process, the concrete method that generates the envelope of multiple groups of original test data on and off the line is: Divide the sampling time into N intervals of equal time, take the maximum value of multiple sets of original test data in each interval, and perform fitting to form an upper line of the envelope; take the minimum value of multiple sets of original test data in each interval, and perform The fit forms an envelope lower line. 6. according to the described method of claim 1, it is characterized in that: in described step (2), multiple groups of original test data are carried out time alignment process, the concrete method that generates the envelope of multiple groups of original test data on and off the line is: Fit each set of original data in multiple sets of original test data to obtain multiple fitting curves; select a set of sampled data as a benchmark, and calculate the value of each point on the multiple fitting curves corresponding to each sampling time of the set of sampled data The maximum value is fitted to form the upper line of the envelope; the minimum value of each point on the multiple fitting curves corresponding to each sampling time of the set of sampling data is calculated, and the fitting is performed to form the lower line of the envelope. 7. The method according to claim 1, characterized in that: the multiple sets of original test data entered in the step (1) select the original test data within one year. 8. The method according to claim 3, characterized in that: if it is judged to be qualified, it also includes step (5) of inputting the test data to be diagnosed in the MySQL database. 9. The method according to claim 1, characterized in that: multiple groups of original test data are greater than 5 groups.