A review of artificial intelligent methods for machined surface roughness prediction

With the increasing demand for ultra-precision machining parts in the industrial field, surface roughness, as a key quality indicator, faces dual challenges of precision and speed in measurement technology. Recent explosive advancements in artificial intelligence (AI) have significantly enhanced the capabilities of intelligent surface roughness prediction. This paper presents a comprehensive review and comparative analysis of artificial intelligence (AI) methods utilized for surface roughness prediction in machining. It systematically examines various AI prediction techniques, encompassing machine learning, image processing, and deep learning, while highlighting their respective advantages and limitations. A thorough assessment is conducted from multiple dimensions, including prediction precision, accuracy, range, adaptability, machining methodologies and materials and other influences. Additionally, future research trends in intelligent surface roughness prediction are outlined based on the current technological landscape. Notably, in terms of algorithmic advances, our review reveals that deep learning methodologies are increasingly preferred for their efficiency. In terms of prediction precision, most research can achieve with acceptable relative error (below 0.1 mu m). We also observe a positive correlation between relative error and prediction range. Furthermore, most studies get significant advancements in achieving minimum roughness detection values below 0.5 mu m, and even reaching 10 nm in some cases. In addition, most studies have certain requirements for the detection environment in terms of methods applicability, such as offline/online settings, light sources, dataset size and computational resources. The applicability of different artificial intelligence methods to different materials also varies. Moreover, a statistical analysis of surface roughness evaluation parameters reveals a predominant focus on Ra, with some studies also utilizing Rz. Lastly, the purpose of this review is to offer a deeper understanding of AI-based surface roughness prediction in machining through comprehensive statistical analyses and critical insights, providing valuable guidance for future research endeavors in this domain.