In response to the periodic abnormal fluctuations in the waste solvent recovery tank data collected in real-time from a paint mixing room in a certain paint workshop,this paper tested different algorithms,including anomaly detection algorithms,conventional filtering algorithms,and multi-algorithm fusion methods,to filter the liquid level curve.The final test results showed that a self-designed multi-algorithm fusion method integrated the advantages of various filtering algorithms and effectively solved the problem of disto rtion in the water-based waste solvent level data,providing a good data foundation for subsequent research on reducing water consumption in waste solvents.At the same time,this method can be applied to related equipment in various paint workshops,relying on big data analysis methods to deal with liquid level disturbance problems and providing high-quality data for subsequent big data analysis research.