Industrial real-time data stream provides great potential in producing big data to support industrial system optimization and intelligent management decision-making. However, storingall massive industrial data without compression requires a huge storage space. Meanwhile,the online transmission consumes a lot of network bandwidth, which brings high pressure to the data transmission line. In practice, the real-time data collected by industry are not always variable and valuable, and contain many redundant data, which obviously affects the efficiency of big data analysis. How to efficiently compress the industrial real-time data, relieve the pressure of industrial real-time data transmission, reduce the storage overhead, and improve the analysis efficiency, is the key technical problem to be solved in industrial manufacturing. To address this problem, the industry has proposed a variety of data compression algorithms, such as revolving door compression algorithm, sliding window compression algorithm and Bayesian compression algorithm. However, there are still some issues such as low compression efficiency, large amount of computation, and poor compression effect. Aiming at the requirement of industrial real-time data compression, this work proposes a nonlinear on-line process data compression algorithm termed NOCAQF based on quadratic function. Simulation and comparative experiments show that the proposed algorithm has the advantages of simple parameters, high compression rate and small amount of computation. When the data collection system processes massive real-time concurrent data streams, NOCAQF can significantly reduce the system load and improve the comprehensive compression rate on the premise of small amount of computation.