Xiaoliang Jin

In Direct Energy Deposition (DED), the melt pool temperature is a critical control parameter thataffects deposition rate, porosity formation, residual stress, and microstructure in the final parts. Inthis thesis, a data-driven approach using Machine Learning (ML) models is used to predict the meltpool temperature using experimental data. This thesis presents the integration of the laser-basedDED system using metal powder feedstock, the determination of the process parameter window forthe setup, and the development of an ML pipeline to predict the melt pool temperature based onits history.In the system integration for the DED system, a laser generator, powder feeder, depositionhead, and sensors (i.e., an IR camera and a 2-wavelength pyrometer) were integrated into an existing3-axis motion stage. Python-based software was developed to control the laser generatorand to read data from the sensors. The software calibrates the IR camera’s temperature, whichis highly dependent on the emissivity, by leveraging the data from the 2-wavelength pyrometer.To determine the process parameter window, 150 single-layer clads were deposited; clads’ crosssectionswere polished and etched, and optical microscopy was used to measure the clad’s height,melt pool’s depth, and dilution ratio. Analysis was conducted on the correlation of the processparameters, laser power, scan speed, flow rate, and the measured properties of the clads. Theprocess parameters with the minimal dilution of (5-25%) were selected to obtain clads with propergeometry and bonding to the substrate.Finally, the temperature data of a 6-layer thin wall with the obtained process parameters wereused to train several ML models, including Dense Neural Networks (DNN), 1-Dimensional ConvolutionalNeural Networks (1D-CNN), Recurrent Neural Networks (RNN), Long Short-Term Memory(LSTM), and Gated Recurrent Unit (GRU). LSTM shows better performance among these models;therefore it was implemented in the ML pipeline for temperature prediction. The Model canpredict the trend and fluctuations of the melt pool temperature with higher accuracy compared tothe existing models for melt pool temperature prediction in the DED process.

View record

Edge trimming is a necessary machining operation to achieve the required dimension and surface quality of carbon fiber reinforced polymer (CFRP) components. Rapid tool wear occurs during the edge trimming process owing to the high strength and abrasive nature of the carbon fibers, leading to potential surface damage. To satisfy the production efficiency and part quality requirements, prediction and monitoring of tool wear progression are essential in edge trimming of CFRP.This thesis presents the tool wear prediction and monitoring in edge trimming of unidirectional (UD) and multidirectional (MD) CFRPs using the artificial neural network (ANN) and random forest (RF) algorithms. The objective is to achieve the tool wear prediction and monitoring under various process conditions with limited training data. For edge trimming of UD CFRP, the feature parameters of the machine learning models consist of the instantaneous process parameters, such as the instantaneous radial force, fiber cutting angle, and uncut chip thickness. As a result, the training data obtained at one radial and axial depth of cut are applicable for tool wear prediction at other radial cutting depths. The experimental data shows a high correlation between the instantaneous radial force and the tool flank wear length. The proposed method achieves the tool wear length prediction within 10% error under various edge trimming conditions. The experimental data from UD CFRP is used for tool wear monitoring in edge trimming of MD CFRP. The tool wear length ratios among the tool edge portions corresponding to different UD layers of the MD CFRP are experimentally identified. The overall radial force is predicted by individual force components from each UD layer, which are influenced by the cutting speed, feed rate, tool helix angle, and wear length distribution along the tool edge. The results show that the proposed model predicts the radial force in edge trimming of MD CFRP with different layer-up sequences within 20% error including the tool wear effect. The proposed method is able to perform tool wear monitoring in edge trimming of MD CFRP based on the force prediction with the predefined maximum tool wear length.

View record

Carbon fiber reinforced polymer (CFRP) composites have been widely used in aerospace, aviation, and automotive industries due to their high strength/stiffness-to-weight ratios, high temperature resistance and corrosion resistance. CFRP components are usually produced in near net-shape, and cutting operations such as drilling, slot milling, and edge trimming are required to remove excessive materials and fulfill the geometry and surface quality requirements of the final parts. Practical cutting operations are in the form of oblique and sequential cutting at the tool edge. Different from metal cutting, the material removal mechanism and surface quality are highly dependent on the fiber orientation in cutting CFRP materials. This thesis presents a 3-D finite element model of oblique cutting of unidirectional CFRP. The effects of the fiber orientation and oblique angles on the chip formation, cutting forces, and subsurface damage are simulated and analyzed. It is found that the out-of-plane force and the depth of subsurface damage increase with the oblique angle in all fiber orientation angles except 0°. To represent the nature of sequential cutting, a second cut on the machined material with existing damages and residual stresses due to previous cutting is simulated. The results show that the effect of sequential cutting on the cutting forces is the largest at 90° fiber orientation angle.Oblique cutting experiments were conducted on unidirectional CFRP. The cutting forces and chip morphology between the simulations and the experimental results were compared. The proposed FE model reveals the effect of oblique angle and sequential cutting on the mechanisms of chip formation and surface generation. The results are able to provide guidance in choosing proper cutting parameters and tool geometries to minimize the subsurface damage and potential delamination corresponding to in actual cutting operations of CFRP composites.

View record

Surface texturing is a manufacturing process to generate periodic geometric patterns on component surfaces in order to achieve certain functions, such as tribological property, adhesion, and wettability. This thesis presents a surface texturing technique using ball-end milling with high feed speed and spindle speed modulation. The ratio between the feedrate and the cutting tool radius is in the range of 0.2-0.4 when the spindle speed is a constant, and a certain amount of workpiece material remains after the cutting process to form the surface texture. A sinusoidal modulation signal is added to the spindle speed command, so the spindle speed becomes time-varying in order to generate different texture profiles based on the modulated frequency and amplitude.The cutting tool kinematics of the surface texturing process are modeled considering the tool tip run-out and deflection due to the cutting forces. Z-map method is used to simulate the geometry of the 3-D surface texture based on the tool tip trajectory. The effects of modulation parameters on tool tip trajectories and surface textures are analyzed. The relationship between the micro features of the surface texture and the process parameters are determined. Surface texturing experiments are conducted based on the proposed technique, and tribology tests are performed on the textured surfaces. It is shown that the textured surfaces present frictional anisotropy, which depends on the process conditions and the modulation parameters of the spindle speed. The proposed technique is able to achieve fast generation of various surface textures without additional instrumentation, and the final texture geometry is controllable based on the presented kinematics model.

View record