Organizational Unit:

George W. Woodruff School of Mechanical Engineering

Permanent Link

https://hdl.handle.net/1853/70728

Parent Organization

Organizational Unit

College of Engineering

Includes Organization(s)

Organizational Unit

Advanced Mechanical Bipedal Experimental Robotics Lab

Organizational Unit

Computational Reactor and Medical Physics Laboratory

Organizational Unit

Cryo Lab

Organizational Unit

Environmentally Conscious Design and Manufacturing

Organizational Unit

Fusion Research Center

Show 4 more

ArchiveSpace Name Record

https://finding-aids.library.gatech.edu/agents/corporate_entities/120

Full item page

Publication Search Results

Now showing 1 - 10 of 76

SMART PARALLEL WAVELET TRANSFORMATIONS FOR EDGE AND FOG DETECTION OF BEARING DEFECTS

(Georgia Institute of Technology, 2021-12-14) Rauby, Pierrick

Rolling Element Bearings (REB) are critical components of a wide range of rotating machines. Identifying and preventing their faults is critical for safe and efficient equipment operation. A variety of condition monitoring techniques have been developed that gather large amounts of data using acoustic or vibration transducers. Further information about the health of an REB can be extracted via time domain trend analysis, and amplitude modulation technics. The frequency domain-specific peaks corresponding to the defects can also be identified directly from the spectrum. Such approaches either provide little insight into the type of defect, are sensitive to noise, and require substantial post-processing. Complicating current fault diagnostic approaches are the ever-increasing size of datasets from different types of sensors that yield non-homogeneous databases and more challenging to execute prognostics for large-scale condition-based maintenance. These difficulties are addressable via approaches that leverage recent developments on microprocessors and system on chip (SoC) enabling more processing power at the sensor level, unloading the cloud from non-used or low information density data. The proposed research addresses these limitations by presenting a new approach for bearing defect detection using a SoC network to perform a wavelet transform calculation. The wavelet transforms enable an improved time- frequency representation and is less sensitive to noise than other classical methods; however, its analysis requires more complex processing techniques that must be executed at the edge (sensor) to limit the need for cloud computing of the results and large-scale data transmission to the cloud. To enable near real-time processing of the data, the BeagleBone AI SoC is employed, the wavelet transforms, and the defect classification are achieved at the edge. The contributions of this work are as follows: first, the real-time data acquisition driver for the SoC is developed. Second, the machine learning algorithm for improving the wavelet transform and the defect identification is implemented. Third federated learning in a network of SoC is formulated and implemented. Finally, the new approach is benchmarked to current approaches in terms of detection accuracy, and sensitivity to defect and was proven to obtain between 80 and 90 percent accuracy depending on the dataset.
Build Plate Conduction Cooling for Thermal Management of Wire Arc Additive Manufactured Components

(Georgia Institute of Technology, 2021-10-22) Heinrich, Lauren E.

Wire arc additive manufacturing (WAAM) is often implemented due to its lower investment cost and equipment availability as compared to other metal additive manufacturing techniques. WAAM adds large amounts of thermal energy to a component during the build process, and depending on the size of the component, the cool-down time can be more significant than the deposition time, decreasing productivity. Conduction cooling applied to the build plate was proposed as a strategy to decrease overall part temperature faster than natural energy dissipation, enabling shorter cool down time with comparable part quality. Based on the challenges presented above, the major research question addressed in this thesis is: How does build plate conduction cooling affect process outcomes of WAAM depositions including process time and quality? This thesis presents the results of conduction cooling applied to the build plate of WAAM components using Gas Metal Arc Welding (GMAW) as the additive process. The present results show that build plate conduction cooling can be used to thermally isolate fixturing, decrease the dwell time by 50%, and decrease the cool down to room temperature time by 75%. Build plate cooling was shown to keep the fixturing at 33% the temperature of the conventional sample, indicating that cooling can be used to thermally isolate the fixturing from the build plate of the deposition. The thermal results of the deposition were also accurately modeled using Finite Element Analysis (FEA) to help determine the maximum temperatures achieved during the depositions. The user can set up the cooling system before the deposition begins and the approach does not require user adjustment throughout the deposition time making this system more user friendly than other processes investigated in literature. The system can also be implemented in retrofit systems without major system modification which decreases cost compared to other thermal isolating techniques.
Ensemble Machine Learning Model Generalizability and its Application to Indirect Tool Condition Monitoring

(Georgia Institute of Technology, 2021-08-24) Schueller, Alexandra Gayle

A practical, accurate, robust, and generalizable system for monitoring tool condition during a machining process would enable advancements in manufacturing process automation, cost reduction, and efficiency improvement. Previously proposed systems using various individual machine learning (ML) models and other analysis techniques have struggled with low generalizability to new machining and environmental conditions, as well as a common reliance on expensive or intrusive sensory equipment which hinders their industry adoption. While ensemble ML techniques offer significant advantages over individual models in terms of performance, overfitting reduction, and generalizability improvement, they have only begun to see limited applications within the field of tool condition monitoring (TCM). To address the research gaps which currently surround TCM system generalizability and optimal ensemble model configuration for this application, nine ML model types, including five heterogeneous and homogeneous ensemble models, are employed for tool wear classification. Sound, spindle power, and axial load signals are utilized through the sensor fusion of practical external and internal machine sensors. This original experimental process data is collected through tool wear experiments using a variety of machining conditions. Four feature selection methods and multiple tool wear classification resolution values are compared for this application, and the performance of the ML models is compared across metrics including k-fold cross validation and leave-one-group-out cross validation. The generalizability of the models to data from unseen experiments and machining conditions is evaluated, and a method of improving the generalizability levels using noisy training data is examined. T-tests are used to measure the significance of model performance differences. The extra-trees ensemble ML method, which had never before been applied to signal-based TCM, shows the best performance of the nine models.
A DEEP LEARNING APPROACH TO ADDITIVE MANUFACTURING PROCESS MONITORING AND CONTROL

(Georgia Institute of Technology, 2021-05-12) Klesmith, Zoe

Currently, determining the layer height command to keep a consistent standoff distance in directed energy deposition (DED) is a time and resource intensive trail-and-error process. Additionally, the same is true for choosing process parameters to prevent over and under deposition for part geometries with small-angled corners and high curvature features. Due to the complexity of the additive manufacturing (AM) process as well as the uniqueness of each part geometry, further development of real-time process monitoring, and control is needed for reliable part dimensional accuracy. Specifically, the use of deep learning (DL) offers a promising solution for real-time process control due to DL’s ability to recognize complex, nonlinear patterns with high accuracy. In this thesis, a set of experiments were designed to train two machine learning models to indirectly detect bead height as well as standoff distance from a coaxial image of the melt pool. The performance of these models was accessed, and it was found that including the laser’s reflection on the nozzle in the image greatly increased prediction accuracy. Overall, the top mean accuracy achieved was 99.74% for standoff distance classification, and the top mean accuracy achieved was 99.87% for bead height classification.
Multi-agent open architecture for process monitoring and part certification

(Georgia Institute of Technology, 2021-02-23) Saleeby, Kyle Scott

Methods to connect manufacturing machines, processes, and sensors have rapidly developed through the fourth industrial revolution, known as Industry 4.0. Data collection is now possible at every point in a production process, providing exceptional analysis opportunities to monitor and affect manufacturing operations. Digital manufacturing technologies can be applied to computer numeric control (CNC) manufacturing processes to measure and improve component quality. Various architectures have been proposed to leverage machine connection mechanisms and extract information in a logical manner. However, these architectures often rely on proprietary software, restricting flexibility for future changes and upgrades. Furthermore, additional capabilities are needed to combine data collected from different sensing modalities and provide a method of in-situ geometric verification. A multi-agent open architecture is proposed to collect, analyze, and communicate information of different formats and sampling characteristics in a strategic manner. This body of work evaluates the strategic combination and synchronization of information from multiple sensing modalities to improve the accuracy of digital twin models. A voxel modelling methodology is developed and investigated to create a digital twin of the component being produced. Information describing the machine’s current operations is strategically combined with information from additional sensing modalities, improving the accuracy of in-situ digital twin models by up to 52%. This research results in (1) a method to geometrically compare features of in-situ components from multiple sensing modalities against desired specifications, (2) a multi-agent architecture to support efficient communication, storage, and use of this information, resulting in (3) feedback methodologies for commercial CNC systems to affect the in-situ manufacturing process and correct geometric deviations.
Bead Geometry, Microstructure and Texture in Multi-Axis Directed Energy Deposition

(Georgia Institute of Technology, 2020-07-30) Elsayed, Omar Hesham

The purpose of this study is to examine the effect of varying laser incidence angles on textural, microstructural and geometric characteristics of DED-processed materials, provide a more comprehensive outlook on participating laser-matter interaction phenomena and ultimately devise strategies to ameliorate print performance. In this study, single-layer, single-/multi-track specimens were processed to examine the effect of non-orthogonal angular configurations on bead morphology, microstructure, phase composition and textural representation of DED-processed materials. Laser power measurements were also conducted to understand the effect(s) of laser spot size changes and laser attenuation on the effective power reaching the substrate surface. It has been observed that bead morphology experienced a normal curve behavior with increasing lead angles, while it experienced a decrease with decreasing lean angles. The orthogonal setting exhibited the most promising bead morphology values. An asymmetry in the distribution of the bead morphology plots indicates that there is preferential bias in utilizing a certain angular configuration over another in terms of potential incorporation of non-orthogonal deposition operations in industrial applications. No significant differences in phase composition, texture and microstructure were observed with processed samples of various angular configurations as well as raw, unprocessed powders, indicating that a potential route for enhanced process robustness is achievable without significantly affecting bulk material properties. In unfocused beams, laser incidence angles have little effect on the extent of laser power loss due to both laser spot size variations as well as powder cloud laser attenuation, indicating that more involved studies are required.
Bayesian edge analytics of machine process and health status in an IoT framework

(Georgia Institute of Technology, 2020-07-21) Newman, Daniel Merle

Using modern machine learning tools and embedded computing, a low-cost, integrated data acquisition platform is proposed in this work. Built on modern, open-source hardware and software, this platform enables high-quality sensor data acquisition and edge-based computation to facilitate machine health monitoring in an IoT framework. By leveraging proposed protocols for edge-based feature extraction, high-volume sensor data payloads are reduced in size to facilitate health monitoring and near real-time inference. The computational latency of this proposed methodology compares favorably to cloud-based solutions, where network transmission latency introduces significant variance in obtaining statistical features and model inference. A case study in tool wear analysis shows that CNC controller data may be used to contextualize accelerometer measurements and, in turn, facilitate training novelty detection and classification algorithms. These algorithms are then deployed to the edge device for near-real time inference.
Surface qualification toolpath optimization for hybrid manufacturing

(Georgia Institute of Technology, 2020-07-21) Thien, Austen Edward

Hybrid manufacturing machine tools have been shown to have great potential in revolutionizing the manufacturing of components by combining both additive manufacturing (AM) and subtractive manufacturing (SM) processes on the same machine tool. However, a prominent issue that can occur when going from AM to SM is that the toolpath for the SM process does not take into account the geometric discrepancies caused by the previous AM step. Thus, the toolpaths used for the SM process are inefficient and can lead to increased production times and increased tool wear, particularly in the case of wire-based directed energy deposition (DED). This work discusses a methodology for approximating the geometric surface of parts manufactured using an on-machine touch probe to gather geometric data and create a digital twin of the part surface. Three different geometric approximation methods using minimal probe points are formed: triangular, trapezoidal, and an augmented hybrid of the two. Optimized SM toolpaths are created using each geometric approximation with multiple objectives of reducing total machining time, surface roughness, and cutting force. Different prioritization scenarios of the multi-objective optimization goals are evaluated to determine efficiency and quality trade-offs. Based on multi-objective optimization results for all prioritization scenarios and a comparison of the toolpaths generated for each geometric approximation, the optimal geometric surface approximation is determined to be the augmented geometric approximation. Furthermore, it is shown that prioritization of the machining time and cutting force optimization goals leads to poor performance improvements in the other optimization goals.
Qualification of laser powder bed fusion processed 17-4 PH stainless steel as a function of powder condition

(Georgia Institute of Technology, 2020-07-21) Berez, Jaime Michael Schnaier

The metal additive manufacturing process of laser powder-bed fusion (LPBF) presents a challenge to develop qualified processes to match the rapid pace of technology development. An aspect of the LPBF process where this applies is defining the how powder feedstock conditions affect the quality of produced components. This study examines how in-machine powder feedstock supplies evolve and are otherwise affected during the LPBF process, and how these effects impact subsequent builds which use said feedstock. An examination of powder flowability, rheology, and morphology is conducted to characterize the powder conditions. To study the effects, an assessment of produced component tensile, fatigue, and microstructural properties is conducted. Fatigue life is analyzed using a reliability modeling approach in order to provide detailed statistical conclusions often missing in other analyses. Powders are found to evolve their characteristics over exposure to repeated LPBF processes, particularly in the extremes of powder size distribution and measures of bulk flow. No significant effects on microstructural, hardness, tensile, and fatigue properties of the produced components are shown. Fatigue life is discovered to exhibit a dependence on spatial origin of the produced component. Additionally, a detailed characterization of the scatter in fatigue life typical to the process and material is provided.
Development and evaluation of interfacial structures for hybrid manufacturing

(Georgia Institute of Technology, 2020-07-14) Feldhausen, Thomas A.

Hybrid manufacturing is a combination of additive (deposition) and subtractive (machining) manufacturing in a single machine tool. Such a system can be used for near net shape manufacturing and component repair using either similar or dissimilar materials. This dissertation investigates methodologies for laser wire-fed hybrid manufacturing processing for commercially available systems and demonstrates how process parameters can be optimized resulting in a deposition rate of 2.5 kg/hr of steel. Integrated into a single system, transition between additive and subtractive manufacturing can occur immediately and be leveraged to generate large components by alternating between the processes. This dissertation investigates how this capability can reduce overall cycle time by up to 68%, improve average elongation to failure by 22%, and reduce average porosity by 16%. With hybrid manufacturing systems, it is now feasible to control the interfacial conditions between the substrate and deposition. Other deposition processes require substrates to be planar, but hybrid manufacturing’s subtractive capability allows for unlimited surface structures and conditions. This dissertation further investigates multiple surface structures for similar and dissimilar materials but concludes that these structures do not result in any improvement of mechanical properties. As a result, these investigations has not only set the foundation for laser wire-fed hybrid manufacturing process development, but has influenced the direction of future research in the field.