Abstract: The discharge of wastewater in petrochemical production is large, and several to tens of tons of wastewater are discharged per ton of chemical products produced. Although domestic petrochemical related enterprises have built sewage treatment equipment and standards according to standards, there are still serious environmental damage and economic losses caused by human negligence, improper disposal, facility defects and other factors that result in non-compliant discharge. This paper relies on collaborative work such as open-source neural networks, visual color grading analysis, and sniffing devices to achieve multi-dimensional recognition and judgment of pre-discharged wastewater. This provides a real-time and controllable channel for petrochemical enterprises to supervise their own discharge situation. Therefore, sewage discharge due to negligence, equipment defects, non-standard operations and other reasons can be eliminated so as to reduce production risks for enterprises.

Key words: petrochemical emissions, sewage, multi-order visual analysis, sniffing analysis, sewage identification