6533b820fe1ef96bd127a130

RESEARCH PRODUCT

Research and implementation of artificial neural networks models for high velocity oxygen fuel thermal spraying

subject

description

In the high velocity oxygen fuel (HVOF) spray process, the coating properties are sensitive to the characteristics of in-flight particles, which are mainly determined by the process parameters. Due to the complex chemical and thermodynamic reactions during the deposition procedure, obtaining a comprehensive multi-physical model or analytical analysis of the HVOF process is still a challenging issue. This study proposes to develop a robust methodology via artificial neural networks (ANN) to solve this problem for the HVOF sprayed NiCr-Cr3C2 coatings under different operating parameters.First, 40 sets of HVOF spray experiments were conducted and the coating properties were tested for analysis and to build up the data set for ANN models. The relationship among the process parameters, behaviors of in-flight particles, and coating properties were investigated from an initial view, which provided a preliminary understanding of the HVOF process and sprayed coatings. Even though the effect of process parameters on the behaviors of in-flight particles and thus on the coating’ properties can be roughly summarized, it is impossible to build up direct connections among them.Second, two ANN models were developed and implemented to predict coating’s performances (in terms of microhardness, porosity and wear rate) and to analyze the influence of operating parameters (stand-off distance, oxygen flow rate, and fuel flow rate) while considering the intermediate variables (temperature and velocity of in-flight particles). A detailed procedure for creating these two ANN models is presented, which encodes the implicitly physical phenomena governing the HVOF process. A set of additional experiments was also conducted to validate the reliability and accuracy of the ANN models. The results show that the developed implicit models can satisfy the prediction requirements. Clarifying the interrelationships between the spraying conditions, behaviors of in-flight particles, and the final coating performances will provide better control of the HVOF sprayed coatings. Additionally, mean impact value (MIV) analysis was conducted to quantitatively explore the relative significance of each input on outputs for improving the effectiveness of the predictions.Lastly, the well-trained ANN models were programmed and integrated into the homemade HVOF spray control system to realize an intelligent control system. With this system, the temperature and velocity of in-flight particles can be calculated by entering process parameters, and thereafter obtaining specific coating properties. A reverse ANN model was also integrated, which calculates process parameters based on the microhardness of the coating to guide the selection of the best parameters. This integration provides a preliminary idea for the construction of an intelligent control system for HVOF spray process and can be promoted to other thermal spray technologies.Overall, based on a large data set, this work not only intuitively analyzed the relationship among process parameters, behaviors of in-flight particles, and coating’s properties, but also provided a prediction method for the HVOF spray process and HVOF sprayed coatings via the optimized and well-trained ANN model. In addition, a prototype to realize an intelligent control system for HVOF spray process has also been suggested.