Sensor self-diagnosis method for a nickel top-blowing furnace based on graph interactive dynamic fusion

Abstract

As modern industry gradually advances towards greater automation and intelligence, the scale of nickel top-blowing furnace smelting systems is continuously expanding, leading to an increasing need for sensor maintenance. Traditional periodic evaluations and manual maintenance methods are no longer sufficient to meet the development needs of intelligent sensors. To address this issue, this paper proposes a sensor self-diagnosis method based on graph interactive dynamic fusion, called DLGCN-GIDF. First, a combination of knowledge-driven and data-driven approaches is introduced. By constructing a dual-layer architecture based on a functional module graph network and a sensor graph network, a sensor correlation graph model for the nickel top-blowing furnace system is established. Next, with the aid of agraph interactive dynamic fusion module, the relative weights between functional modules and sensors are integrated to perform spatiotemporal correlation-based graph fusion. This enables the prediction of spatiotemporal data for sensors from a system perspective. Finally, the goal of sensor self diagnosis is achieved using a standardi- sed residual testing algorithm. Takinga nickel top-blowing furnace smelting system as an example, the feasibility and effectiveness of ourmethod ofsensor fault self-diagnosis are verified.