Laser cladding state recognition and crack defect diagnosis by acoustic emission signal and neural network

Highlights

• Monitor the laser cladding process using acoustic emission detection technology.

• More comprehensive technological parameters affecting crack activity are studied.

• The abnormal process parameters in laser cladding process were detected.

• Neural network is used for signal recognition and diagnosis.

Abstract

Laser cladding technology uses a high-power laser beam to melt the substrate and metal powder at high temperature to form a molten pool. Relying on the spontaneous cooling of the molten pool, a metal cladding coating is formed on the substrate to strengthen the surface properties of the substrate metal. However, the typical defects such as cracks are easy to occur during the cladding process, which greatly affects the performance and quality of the cladded layer. This paper proposes a method for the state identification of cladding and the crack detection in the laser cladding process. By monitoring the acoustic emission signal during the laser cladding process, the current cladding state such as the status of laser power, scanning speed, and powder feed rate, and the occurrence of cracks are identified. By collecting the acoustic emission signal, the method first performs the data preprocessing for signal feature components according to the characteristic parameters of the signal maximum peak value and the energy of the emission signal samples, and then a deep learning neural network is applied to extract the feature vectors based on the two major characteristics of the signal. Finally, the current cladding states are recognized and the generation of cracks are detected based on the extracted feature vector and the identification through the neural network.

Introduction

With the continuous development, laser cladding technology possesses the many advantages in manufacturing of large and complex metal workpiece and in processes of metal surface strengthening. Laser cladding technology is a technical process to melt the substrate and metal powder by high-power laser beam to form a new layer of melt on the substrate. The pool relies on spontaneous cooling to form a cladding layer with high hardness, good thermal stability and certain metallurgical, mechanical or physical properties on the surface of the base material [1], thereby significantly improving the wear resistance, corrosion resistance, and resistance of the metal material fatigue and high temperature resistance [2], [3]. However, because the laser cladding process is a complex process with multi-physics integration, multiple process parameters interaction among laser power, laser scanning speed, powder feed rate, etc., which will result in an interactive effect on the quality of the cladding metal layer [4]. The coating is prone to severe cladding defects such as cracks, which greatly limits the wider promotion and development of laser cladding technology [5]. The optimization of the cladding process parameter combination and the feedback adjustments to the monitoring of the molten pool state of the cladding process can greatly improve the quality of the cladding layer and reduce the cladding defects such as cracks [6].

Therefore, the research on the identification of cladding state during laser cladding and the detection of typical defects in cladding layers have become research hotspots at home and abroad in recent years. In fact, the identification of cladding state and online defect detection in laser cladding have also been a challenge in the field of metal additive manufacturing, there are still few research reports on this aspect [7]. For the detection of cladding layer defects during the laser cladding process, Haythem Gaja [8] of the US Missouri University of Technology used cladding powders with different ratios to cladding to artificially control the occurrence of crack defects or pore defects, and construct the artificial neural network model to identify the acoustic emission signals collected during the process. The experiment can successfully identify these two types of defects, indicating that the acoustic emission signals have great potential for defect detection during the laser cladding process. Barua [9] et al. proposed the RGB value detection method of the laser metal deposition (LMD) high-temperature melt channel image. In large-spot LMD additive manufacturing process, the camera was used to shoot the red hot melt channel, and the RGB value of the melt channel image was analyzed online. If abnormal highlight areas were found, it is inferred that there are large crack defects under this area, but this method is limited to identify big cracks. For the related research on the quality evaluation and identification of the cladding layer, for example, Sashevchik et al. [10] adjusted the process parameters to produce pores with different densities, and defined different quality grades with different porosities. The sound signal in the cladding process is recorded and classified by machine learning algorithms. The experiment obtains a good accuracy but the method of diagnosis on quality level, does not provide any adjustment for laser cladding equipment information, to improve the quality of cladding layer to provide timely, effective contribution. Ye [11] changed the process parameters to make the parts in classes of “good”, “medium”, “poor” and other quality levels, and built a quality classification model, based on deep confidence network (DBN), to analyze the process acoustic sound signal, which experiment achieved a higher accuracy. However, only two process parameters, scanning speed and laser power, were changed to obtain several different cladding qualities, and the influence of other process parameters on cladding quality was not involved. Hossein Taheri et al. [12] detected acoustic emission signals from cladding substrates, and defined five states of the process: machine in inactive state, powder feeding only on, normal cladding state, low laser power, and low powder feeding rate. The clustering analysis of acoustic emission signals show that the data of each category has better clustering effect. His work demonstrates the application potential of AE in the on-line monitoring of laser cladding technology, but there are no results on the effect of more process parameters on the cracks in the cladding layer.

In order to monitor the parameters’ changing of laser cladding process online and detect the generation of typical defect cracks in cladding layer, in this research study, the acoustic emission (AE) signal feature-vectors are extracted, based on the characteristic of the maximum magnitude and power energy of the signal, to extract the high-energy signal samples and the low-energy signal samples, which containing the high-energy impact signals and the low-energy base signal respectively. Then, a neural network for crack detection is applied on the high-energy signal samples and detect whether cracks are generated from the additive process of laser cladding. A neural network for identifying cladding state is applied on the low-energy signal samples and identify the current cladding state. Finally, according to the feature-vector extracted from these two parts of the data, the current state of the process is recognized as one state of the following: the state of normal cladding with good cladding layer quality; the state of a certain process parameter is abnormal and possible cracks in cladding layer; or the state of a certain process parameter is abnormal and the detected cracks.