Vision-based defect detection in laser metal deposition process

Abstract

Purpose – Laser metal deposition (LMD) is a type of additive manufacturing process in which the laser is used to create a melt pool on a substrate to which metal powder is added. The powder is melted within the melt pool and solidified to form a deposited track. These deposited tracks may contain porosities or cracks which affect the functionality of the part. When these defects go undetected, they may cause failure of the part or below par performance in their applications. An on demand vision system is required to detect defects in the track as and when they are formed. This is especially crucial in LMD applications as the part being repaired is typically expensive. Using a defect detection system, it is possible to complete the LMD process in one run, thus minimizing cost. The purpose of this paper is to summarize the research on a low-cost vision system to study the deposition process and detect any thermal abnormalities which might signify the presence of a defect. Design/methodology/approach – During the LMD process, the track of deposited material behind the laser is incandescent due to heating by the laser; also, there is radiant heat distribution and flow on the surfaces of the track. An SLR camera is used to obtain images of the deposited track behind the melt pool. Using calibrated RGB values and radiant surface temperature, it is possible to approximate the temperature of each pixel in the image. The deposited track loses heat gradually through conduction, convection and radiation. A defect-free deposit should show a gradual decrease in temperature which enables the authors to obtain a reference cooling curve using standard deposition parameters. A defect, such as a crack or porosity, leads to an increase in temperature around the defective region due to interruption of heat flow. This leads to deviation from the reference cooling curve which alerts the authors to the presence of a defect. Findings – The temperature gradient was obtained across the deposited track during LMD. Linear least squares curve fitting was performed and residual values were calculated between experimental temperature values and line of best fit. Porosity defects and cracks were simulated on the substrate during LMD and irregularities in the temperature gradients were used to develop a defect detection model. Originality/value – Previous approaches to defect detection in LMD typically concentrate on the melt pool temperature and dimensions. Due to the dynamic and violent nature of the melt pool, consistent and reliable defect detection is difficult. An alternative method of defect detection is discussed which does not involve the melt pool and therefore presents a novel method of detecting a defect in LMD.