Abstract: Carrying out compliance checking of old equipment design documents is conducive to accurately controlling the risks accumulated in the development process of the refining and chemical industry. It is an inevitable requirement for overall development and safety. In this paper, the text data in the static equipment design documents of the old device is studied by machine learning algorithm, and the data analysis model is constructed, aiming to obtain and analyze the key information of the old device that does not meet the compliance of the design documents. The research mainly consists of three aspects: the first is to use data preprocessing method for data cleaning; the second is to use TextRank algorithm to achieve automatic acquisition of the keywords and key phrases in a large number of texts; the third is to use LDA algorithm to train the topic model to automatically generate the key topics in the text. The results obtained from this method can assist experts in rapid and accurate investigation and assessment of safety risk information of old equipment.

Key words: compliance checking, machine learning, data cleaning, TextRank algorithm, LDA topic model