Online prediction of grinding wheel condition and surface roughness for the fused silica ceramic composite material based on the monitored power signal

• Published in: Journal of materials research and technology Vol. 24; pp. 8053 - 8064
• Main Authors: Wang, Jinling, Tian, Yebing, Zhang, Kun, Liu, Yanhou, Cong, Jianchen
• Format: Journal Article
• Language: English
• Published: Elsevier B.V 01.05.2023; Elsevier
• Subjects: Fused silica ceramic matrix composite; Grinding power monitoring; Online prediction; Warm-cold SGE density; Warm-cold SGE density; Grinding power monitoring; Online prediction; Fused silica ceramic matrix composite
• ISSN: 2238-7854
• DOI: 10.1016/j.jmrt.2023.05.040

The fused silica ceramic matrix composite (SiO2f/SiO2) is characterized as one of the most promising materials in radar radomes and wave-transparent antenna windows. However, the ceramic matrix surface is rather rough, which reduces their service life in severe environments. Although precision grinding is used to improve surface quality, it gets difficult to precisely achieve the desired surface requirement due to the SiO2f/SiO2 material hardness, brittleness, and complicated fiber structure, as well as the uncertain wheel conditions. Especially the wheel condition and its influence on surface quality are unable to quantify directly. With the fast development of IoT monitoring and cyber-physical system, online prediction using process physics signals gains more interest. Power monitoring is a convenient way to obtain valuable grinding information at a relatively lower cost, making it more popular in the industry. However, the links between grinding parameters, power signal, and wheel conditions, surface quality is unclear. A series of grinding experiments of the SiO2f/SiO2 workpiece has been designed to demonstrate how the wheel conditions are indicated using the warm-cold density chart of specific grinding energy. A novel exponential polynomial model of surface roughness associated with wheel conditions indicated by the cutting power signal has been built. Compared with the conventional prediction model, REs center, MSE, and Pearson correlation coefficient are improved significantly from −0.0213, 0.029669, and 0.8695 to 0.0022, 0.022478, and 0.8710, respectively.