The RAMS (Reliability,Availability,Maintainability and Safety) information of rail vehicles plays an important role in the operation and maintenance of rail transit. It is the key to improve the performance and reliability of railway vehicles.  At present, there are a large number of colloquial, hybridized and subjective data records in the RAMS information of rail transit trains; especially in the vehicle fault record, this phenomenon is widespread, which brings great difficulties to subsequent data analysis. Therefore, how to convert these irregular fault records into computer-readable fault texts is of great significance. Therefore, converting these irregular failure records into  well-structured data is of great significance. This paper proposes a RAMS information structuring framework based on the Jieba word segmentation and Doc2Vec technology. Since Chinese is very different from Indo-European languages such as English and French, spaces cannot be used as separators between words, so Chinese natural language processing has special word segmentation problems. It can be said, however, that due to the complexity of Chinese, there is still no method that can achieve 100% correct Chinese word segmentation. The stagnant word segmentation method used in this article is a statistical word segmentation method. Its basic principle is: based on the prefix dictionary to achieve efficient word graph scanning, generating a directed acyclic graph composed of all possible word formation of Chinese characters in a sentence (Directed Acyclic Graph, DAG). Then use dynamic programming to find the maximum probability path and find the maximum segmentation combination based on word frequency. The so-called prefix dictionary refers to the search of words based on the first and second words of Chinese words as an index, which is exactly the same as the reason we encounter the dictionary in our daily life. In word segmentation, vectorization of text is also a necessary step: we need to convert words into vectors in order to perform subsequent machine learning. The last step is the text similarity calculation. By calculating the text similarity between the pending fault sentence and all standard fault sentences, the standard fault sentence with the highest similarity is selected as the final processing result. The structured data is then compiled into an expert dictionary, describing the railway vehicle information, and a standard failure library which includes all types of possible failures that can happen during the normal operation.  Furthermore, a case study using the failure records from Nanjing Metro and Guangzhou Metro is carried out to test the effectiveness of the proposed framework. It is verified that the expert dictionary and the standard failure library have great impact on the analysis of railway operation data.  Future work on improving the current framework is discussed at the end of this paper. Through the above research, a standard fault sentence library can be finally constructed, which is a refined and summarized fault mode description. Comparing the fault descriptions recorded by frontline maintenance personnel, standard fault sentences are more standard and scientific, and there is only one standard fault sentence in a fault mode. Standard fault statements realize a one-to-one correspondence between fault modes and fault descriptions, and solve the problem that different maintenance personnel have different fault description modes for the same fault mode. Theoretically, the standard fault sentence library should cover all fault modes, but in fact, there are always unknown and uncovered fault modes. Therefore, the standard fault statement library is not perfect, but it should be updateable: when a new fault mode appears, we need to compile a corresponding new standard fault statement, and then add it to the library.


Analysis on RAMS Information for Metro Vehicles Using Natural Language Processing Algorithm: Evidence From China

Division / Sponsors: 
Aerospace, Bioengineering, Applied Mechanics, Design Engineering Division, Advanced Energy Systems, Manufacturing Engineering Division, Noise Control and Acoustics, Nondestructive Evaluation, Diagnosis, and Prognosis (NDPD), Nuclear Engineering Division, Process Industries Division, Fluids Engineering, Heat Transfer, Materials, Microelectromechanical Systems, Safety Engineering and Risk Analysis, Computers & Information in Engineering


Paper Proceedings: https://asmedigitalcollection.asme.org/IMECE/IMECE2020/volume/84478

IMECE2020 - Virtual Conference

Supporting the need among researchers worldwide for additional online and multimedia content, ASME now offers the ASME Conference Video Collection. This platform will collect and provide long-term access to conference presentation videos, encompassing all technology areas within ASME's technical conference and events program. About 30,000-40,000 engineers across all mechanical engineering and related disciplines participate in ASME conferences each year. This includes authors of approximately 6,000 to 8,000 technical papers published by ASME, the majority of whom also submit video presentations for our virtual conferences. The video collection will be available as an add-on to existing product packages or as a separate subscription.