Multi-source heterogeneous big data cleaning method and system in aircraft manufacturing process

Abstract

The invention provides a multi-source heterogeneous big data cleaning method and system in an aircraft manufacturing process. The method comprises the steps that original signals generated in the aircraft part manufacturing process and the assembly process are acquired and subjected to noise reduction preprocessing; performing data diagnosis on the original signal, fitting correlation of a data time sequence by using an LSTM-based model, and discriminating default data and error data by using a sliding window confidence interval strategy to obtain abnormal data; for the abnormal data, repairing and interpolating the data of the abnormal value signal under the established graph model of the Bayesian network according to the graph theory; and judging whether the repaired and interpolated data has quality problems except the basic problems, if so, performing data diagnosis and data repair again, and if not, outputting. The method has the beneficial effects that the data cleaning efficiency can be improved, the value density of a large amount of data generated in the manufacturing and assembling process is effectively improved, and the industrial production benefits are improved.

Description

technical field [0001] The invention mainly relates to the field of aerospace product manufacturing industrial data, in particular to a method and system for cleaning multi-source heterogeneous big data in an aircraft manufacturing process. Background technique [0002] In the actual production of aerospace industry, a large number of sensors are often arranged to collect relevant production data on a regular basis, resulting in many characteristics such as many sources, high collection frequency, large quantity, high dimension, and low quality. When an industrial-related enterprise establishes an information system, even with good planning and design, it cannot guarantee that the quality of the processed and stored data can meet the requirements. Entry errors, data migration, poorly enforced specifications, sensor failures, improper operation, passage of time, and more can all affect the quality of the data stored. Therefore, whether it is to enhance the value content of i...

AI Technical Summary

Problems solved by technology

[0005] (1) The application scenario needs to be more idealized, so it is relatively limited, and it is difficult to face the real-world aircraft manufacturing industrial data

[0006] (2) The data objects of the application are strictly required, and the ability to deal with complex problems is insufficient

[0007] (3) Some methods still require a lot of manpower for manual cleaning and later data integration

Images