Abstract: Currently, massive data has become the core digital foundation for equipment manufacturing enterprises.Due to its characteristics such as multi-source heterogeneity, physical correlation, and spatiotemporal coupling, traditional data detection methods struggle to identify hidden logical conflicts and system-level anomalies.To address the quality detection challenges in complex data scenarios, driven by multi-source data, an intelligent detection method for massive industry data quality based on mixture of experts is proposed.An expert sub-library including regression analysis, clustering analysis, and semantic norms is constructed, and a self-adaptive routing mechanism based on gated networks is designed by combining the Top-k sparse activation routing strategy and the load balancing loss function.Experimental results show that this method has increased the F1 score of mixed data detection from 0.320 0 to 0.705 0, with a routing hit rate of Hit@1 reaching 0.973 7, and overcomes the performance limitations of a single detection model when processing mixed fields.