A set of safety supervision platform for crane operation status based on industrial big data needs emerging information technologies such as complex networking, industrial big data and cloud computing. The platform should cover the whole process of crane, the whole link and the whole life cycle data link. The overall framework of the platform is shown in the following figure.

According to the frame composition of the platform and its key technology, it is necessary to solve the difficulties in the construction process, such as the collection method of complex information of crane operation state, the industrial big data preprocessing method of crane, the mining and analysis technology of crane big data, and so on.

1) Research on data source analysis and field acquisition technology

The crane shall include cranes for hoisting molten metal, cranes for hoisting dangerous goods, large cranes with monitoring systems, explosion-proof environmental cranes and production equipment lifting equipment of key enterprises, etc., which shall be classified according to their respective operating environment characteristics, use characteristics and their own structural characteristics, etc. The operation state information collection scheme of uncertain data source crane suitable for this system is worked out, and the optimal arrangement method of multi-sensor monitoring is studied. The data sources of big data's supervision platform can be divided into the direct upload of crane monitoring software at the field application level and the uploading of the equipment operation and maintenance platform at the enterprise application level. Two types of data are required to unify the standard of two types of data sources and design the data interface.

2) Research on big data preprocessing and Storage Technology of Crane big data preprocessing refers to the use of certain strategies to process the collected crane big data, so that the data can meet the requirements of integrity, consistency, effectiveness and so on, in order to improve the correctness of subsequent data analysis. According to the characteristics of crane site, the data preprocessing of this study mainly lies in missing value processing and similar repeated record processing. Missing value processing is due to the complexity of the field environment and network transmission failures, resulting in the accuracy of some fields in the data set, which is to be processed by statistical methods, classification methods and association rules. Similar repetition The recording is due to the fact that the collected data set may contain repeated records of table unified physical meaning, and the similar detection strategy based on Euclidean distance, Euclidean distance and Pearson correlation coefficient can be used in the processing.

3) Research on data analysis technology and mining method of crane

The purpose of analyzing and mining the collected data is to supervise the crane effectively from many angles and entry points, and to form a supervision mode of the whole life cycle. The common industrial big data analysis and mining methods, such as K-means, BP neural network, genetic algorithm, clustering algorithm and Bayesian theory, are compared and studied, so as to form a unified big data mining algorithm library, which integrates the data mining algorithms corresponding to different business application purposes in the crane supervision platform. Through the application of the data mining method library, the related key characteristics of lifting equipment risk are obtained, and the running state data of lifting equipment is inputted. Platform to automatically obtain equipment daily operation trends and maintenance recommendations.

4) Big data supervision platform architecture and software implementation for multi-type users.Because of the large amount of crane data collected by the platform, it is necessary to adopt a platform architecture with high reliability and fault tolerance, and it needs to have the characteristics of relatively simple programming model. Hadoop model architecture can be adopted. Hadoop is an open source and extensible distributed system architecture, which can realize the distributed file system HDFS, to manage storage resources across clusters. Suitable for this system to face multi-type users based on industrial big data crane safety supervision application. At the same time, considering the software implementation of the platform, ApacheHUE can be used to realize a visual interactive analysis interface, and the different types of the platform can be used. Households can realize the functions of data cluster management, visual display of data analysis results, data cluster status display and so on through Web.