Online detection of powder spatters in the additive manufacturing process

With high power or high brightness laser as a heat source, metal additive manufacturing technology has developed rapidly in recent years. Although this technology is promising in manufacturing, quality detection has always been the main obstacle to its wide application. Traditional offline detection methods are expensive and time-consuming, and therefore cannot be used for online detection. An improved image processing algorithm is proposed to extract the number of powder spatters with a completely melted state from the images under a complex background, can retain the pixel features to the greatest extent, and exhibit high accuracy. To achieve automatic image annotation and predict the forming quality early in the process of additive manufacturing, the 3 Sigma quality evaluation standard on the basis of porosity is established which can solve the problem of single and broad quality evaluation threshold and categorize the quality into four levels. The inverse distance weights K-NearestNeighbor algorithm is then proposed to solve the problem of imbalance between positive and negative samples of different quality classes after quality classification. In general, the classification accuracy of this algorithm is above 95% in all data sets. Compared with other machine learning algorithms, this algorithm has strong interpretability, fewer parameters, faster operation speed, and greater flexibility in online detection. Online detection method proposed in this paper can mine the image information to reveal correlations between process parameters and quality features of interest, thereby monitoring process quality in real-time regardless of offline analysis.