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 16

Open source CNC control with CAM and digital twin integration

(Georgia Institute of Technology, 2019-12-02) Williams, Kyle

High bandwidth internet connectivity and ubiquitous computation are poised to enable automated quality assurance, high efficiency predictive maintenance and an integrated logistic support infrastructure for modern manufacturing. Information technology is in the process of revolutionizing production, as it has revolutionized so many other industries. However, old and new CNC systems alike are unable to fully claim this advantage. Milling machines are a significant capital investment; it is impractical to regularly replace them; aging systems continue to see use, but are increasingly unable to meet modern demands. These demands include tighter machining tolerances, three and five axis automation, and internet connectivity. On the other hand, modern machines evolved in a niche market with a high price for entry; these systems meet performance demands, but employ obfuscated, proprietary hardware/software systems that stifle free market innovation and offer limited bandwidth communication interfaces. They are often prohibitively expensive as well. In this body of work, an aging CNC mill is upgraded with a modern electrical power system and an open source firmware/software architecture for control and communication. A digital twin of the machine tool is developed directly in the CAM environment, where toolpaths are generated. Leveraging this open platform, the CAM software is connected directly to the machine tool over the internet, enabling remote monitoring and control. This report presents the engineering behind the system, in the broader context of the need for open source control and the demands on modern machine tools. The system is vetted out on a 1986 Mori Seiki vertical milling station and experimentally verified.
Automated real-time machine learning for IOT for manufacturing a cloud architecture and API

(Georgia Institute of Technology, 2019-11-12) Parto Dezfouli, Mahmoud

Due to the recent movements in Industry 4.0 and Internet of Things (IoT), accessing or generating data in the Smart Manufacturing (SM) domain has become more attainable; communication protocols such as MTConnect and OPC-UA provide access to a majority of raw data generated from machine tools while retrofit sensor packs facilitate high- frequency data acquisitions from legacy and modern equipment. These technologies have led to the generation of quantities of raw data, known as Big Data (BD), that are complex to be analyzed. Current IoT architectures and frameworks propose Cloud Computing (CC) and Centralized Training (CT) as the addressing solutions for BD and collaborative Machine Learning (ML) models. These solutions, however, have limitations such as Internet dependency and requiring expensive and high-performance cloud resources. As more data are generated, a higher performance framework is required for cloud computing of larger datasets that are either historical in nature or generated from an ever-increasing ubiquitous sensors and sensor arrays that are deployed in modern manufacturing operations. Studying IoT architectures and stream analytics is essential for creation of successful IoT platforms. In this regard, this study proposes a novel, high-performance, and data- driven IoT architecture that considers automated and scalable machine learning techniques with the focus of process control and deeper understanding of manufacturing process and systems performance in the Cyber-Physical Systems (CPS) domain. In this dissertation, first, a novel generalized three-layer IoT architecture utilizing Edge Computing (EC), Fog Computing (FC), CC, and Federated Learning (FL) is presented, where data are preprocessed in the Edge layer, ML models are incrementally trained in the Fog layer and the resulting elements of training are aggregated in the centralized cloud models. Second, two novel stream analytics engines of Outlier Detection and Bayesian Classification, capable of real-time (RT) training and prediction are proposed and analyzed for this architecture. Results show that the training latency for both the Outlier and the Bayesian engines as well as their FL algorithms remained constant as the number of data points increased. On a 1000 data point dataset, the training performances for an upcoming data point for the Outlier and Bayesian engines were on average 136 and 48 times faster, respectively, than retraining the models with all of the data points. These results suggest that the methods discussed in the proposed architecture can lead to the development of higher performance and more scalable IoT frameworks that require lower storage and computing power.
Slicing of tessellated models for additive manufacturing based on variable thickness layers

(Georgia Institute of Technology, 2019-11-12) Han, Dongmin

In contrast to machining or subtractive technologies, Additive Manufacturing (AM) is a set of technologies that fabricate a 3-D object by automatically adding material layer-by-layer. In AM systems, the Computer Aided Design (CAD) model is converted into layers in a process known as slicing. One of the limitations of AM is the geometrical inaccuracy and undesirable surface finish due to the layer-upon-layer application of material. Inclined features suffer significantly from this drawback, known as the stair-step effect. While decreasing the layer thickness can reduce the stair-step effect, the cost of considerably increasing processing time is unappealing to manufacturers. Flat layer additive manufacturing has a number of limitations: first, the trade-off between better surface finish and printing time; second, support structure are usually needed, which causes the unwelcome surface quality on the contact areas between the part and the support structure; and third, the use of flat layer leads to the anisotropy property, which affects the strength of the final parts. To overcome the limitations present in flat layer additive manufacturing, adaptive slicing and curved-layer slicing were proposed. Some research has focused on developing the algorithms that adaptively chooses the layer thickness based on the curvature and angle along the surface. Some has developed curved layer slicing by offsetting the top surface to generate the layers. But all of these works only apply to 3D models with lots of constrains and involve certain level of manual interventions. In addition, all of these works are only aiming at 3-axis FDM machines, while the proposed slicing procedure is not only applicable to 3-axis systems, but also suitable for 5-axis. This research proposes a new solution to slice tessellated CAD models with dynamic thickness layers. The proposed method negates the stair-step effect and provides smooth bonding between layers. It also provide the potential to be applied on 5-axis FDM machines with minimum modifications. In this procedure, the CAD models is divided into planar-curved regions and uniform slicing regions by the directions of the facet vectors. The top surface is extracted from the curved region and the facets are offset with different distance from the top surface to create slicing layers. As a result, the dimensional accuracy is improved using fewer layers compared to uniform slicing. Hence, the proposed method can significantly save print time without compromising quality. In addition, a more generic slicing procedure will be developed by applying the method locally to individual features on a single part. The contributions of the research are as follows: first, a dynamic thickness curved layer slicing algorithm for tessellated models was developed; second, this approach was implemented on a 3-axes Fused Deposition Modeling (FDM) platform; third, the surface integrity property was improved; and fourth, a more generic slicing algorithm was developed for more complicated models.
Developing a device for automatic monitoring of rolling element bearing conditions

(Georgia Institute of Technology, 2019-07-15) Tritschler, Niklas Benedikt

In most instances, rotating machines have a unique vibration signature that relates to their health status. Therefore, vibration analysis is a powerful tool for predictive maintenance. This is especially true for bearings that are a frequent cause of machine breakdown. Presently, bearing analysis of many machines results in significant cost and complexity due to a large amount of vibration data that must be analyzed. The purpose of this thesis is to develop a vibration analysis system that locally collects vibration data, analyzes it automatically and provides feedback as to the bearing condition.
Detection of bearing defects with approximate bearing configuration

(Georgia Institute of Technology, 2019-07-10) Prevost, Eymard Antoine Marie Edudes

Unscheduled maintenance in a production line due to breakdowns is highly detrimental. The ability to predict impending failure and anticipate it is a high value proposition. Such a prediction can be achieved by monitoring components that are known to fail often in mechanical systems, such as bearings. Prior research has led to the development of bearing monitoring approaches widely used today. However, one of the main challenges is the fact that there is often incomplete information about the systems. This thesis will focus on approaches that can be employed to detect bearing defects and incipient bearing failure in the presence of incomplete and inaccurate system knowledge.
A voxelized framework for simulating cutting tool and workpiece interaction

(Georgia Institute of Technology, 2019-04-30) Miers, John Carter

In this manuscript, a voxel based model for the interaction between cutting teeth of an arbitrary end mill geometry and a workpiece is developed that allows for the virtual machining of workpiece volumes with generated tool geometry. In this framework, the workpiece geometry is modeled using a voxelized representation that is dynamically updated as material is locally removed by each tooth of the cutting tool. A ray casting approach is then used to mimic the process of the cutting faces of the tool raking out workpiece material. This ray casting regime is also used to calculate the instantaneous undeformed chip thickness. The resulting voxel based model framework was validated by comparison of predictions with experimentally measured milling forces. The results demonstrate the model’s ability to accurately simulate the interaction of cutting teeth with the bulk material of the workpiece. The model is further expanded to simulate the impact of previous tool passes on subsequent one. Implications of this new voxel based model framework are briefly discussed in terms of utility for predicting local surface finish and computational scalability of complex cutting configurations.
Development of a low-cost wireless accelerometer sensor platform (WASP) for machine monitoring applications

(Georgia Institute of Technology, 2019-04-26) Saleeby, Kyle

The modern Industrial revolution, or Industry 4.0, has dramatically expanded the capabilities of digital manufacturing. However, modern machines with monitoring capabilities are extremely expensive to purchase and take years of operation to recoup the capital cost. A need exists to provide a low-cost Internet-of-Things for Manufacturing (IoT4MFG) sensor platform that can provide accurate monitoring and analysis capabilities on a machine of any age. The Wireless Accelerometer Sensor Platform (WASP) is an extremely low-cost, wireless, and robust solution to upgrade the monitoring capabilities of manufacturing machines to a modern standard. This platform provides a flexible capability to modularly handle analog and digital sensors of standard communication protocols, as well as a standard set of base sensors including an accelerometer. Additionally, optimal placement of accelerometers on a machine is important for the proper measurement of vibrations. Commercial sensors are commonly fastened to the machine through permanent means without proper verification of positional vibration acquisition. A need exists to verify proper placement and function of these sensors. The WASP implements a live vibration monitor with a mobile phone app through Bluetooth Low Energy communication. Combined with a semi-permanent magnetic mount design, the sensors platform allows for precise placement and convenient adjustment to ensure optimal vibration measurement. The design methodology and verification process pursued to develop the WASP are presented. A case study was completed to demonstrate the technology on a manual lathe where machine vibration was measured. The WASP accelerometer was evaluated for accuracy with a spectrum of known input frequencies and results were compared against a high quality baseline accelerometer.
Direct servo control of positional drivatives for 5-axis CNC machine tools using densely-sampled toolpaths

(Georgia Institute of Technology, 2019-03-25) Lynn, David Roby

Ever-increasing quality and complexity requirements for machined parts have led to the development of computer-numerical control (CNC) machine tools with high numbers of servo axes capable of tightly coordinated motion. These machine tools are usually programmed using computer-aided manufacturing software that creates toolpaths for machining surfaces and features selected by a user. Voxel-based computer-aided manufacturing (CAM) software has shown great potential in both creating machining plans for highly complex parts and performing realistic simulations of material removal that would be impractical with current industrial CAM systems. Voxel models allow for the creation of toolpaths that follow the exact surface of a given part on a voxel-by-voxel basis, which enables the recreation of very fine surface details on a machined part. The created toolpaths are translated by the CAM system into a format readable by the machine, known as G-Code, which consists of points and maximum velocities that the machine should follow in order to trace out the desired path. For toolpaths created from a voxel model, this G-Code program consists of many small linear movements for each axis of the machine tool. Specifying toolpaths to the machine in G-Code has a number of limitations: first, many commands are machine specific, which causes compatibility issues between the CAM system and the CNC; second, translating a toolpath into G-Code causes a loss of valuable process control data between the CAM system and the CNC; and third, the use of G-Code forces the CNC to spend valuable compute cycles performing online trajectory planning using a worst-case approach that can prevent the cutting tool from reaching its programmed maximum velocity. Even the most sophisticated CNC machine tool control systems are unable to maintain the programmed tool velocity while machining a toolpath created from a complex voxel model. This causes the machine to not execute the exact toolpath provided by the CAM system, which renders offline simulations of machining and material removal less effective. Much research has focused on finding optimal tool velocities to traverse a path more quickly in order to reduce machining time, but all of these works still rely on G-Code. To overcome the limitations present in G-Code programming, this research develops and evaluates a new solution to offline trajectory planning and control that is enables a CNC machine tool to follow a densely-sampled toolpath (such as one created from a voxel model) at the kinematic limits of each axis. Additionally, the proposed approach will allow for the communication of densely-sampled motion trajectories that would be impossible with standard G-Code. The contributions of this work are as follows: first, a generalized framework and accompanying control system for direct transmission of dense data to and from the machine tool’s servo controllers directly from a voxel-based CAM system is developed; second, a reference implementation of this approach is performed on an open-source CNC platform known as Machinekit; third, near-realtime simulation and analysis capabilities from within the CAM system are developed and discussed; and fourth, the accuracy of motion realizable by the new control system is validated using complex toolpaths created from the CAM system and performance is compared to the standard G-Code programming method.
Extruder dynamics and control in large scale additive manufacturing

(Georgia Institute of Technology, 2018-12-04) Silberglied, Chelsea

In Large Scale Additive Manufacturing, there are many part defects, such as bulging at corners and improper part seams that arise due to a lack of information about the extruder. To reduce the number of part defects, it is necessary to understand and control the extruder dynamics. A bead characterization system (BCS) was created to measure the flow rate out of the nozzle. Tests were run to excite the dynamics of the extruder and perform system identification. Models were then created to describe the system and predict the flow rate out of the nozzle. Then a feed forward controller was implemented to maintain a consistent bead geometry, and the results from the controller were evaluated.
A secure MTConnect compatible IoT platform for machine monitoring through integration of fog computing, cloud computing, and communication protocols

(Georgia Institute of Technology, 2017-12-14) Parto Dezfouli, Mahmoud

Industry 4.0 aims at utilizing advancing computational frameworks such as Human-Computer Interaction and Machine Learning in traditional areas of manufacturing and production. Smart Manufacturing (SM) involves creating intelligent manufacturing systems that use the concepts of Industry 4.0 throughout a product development lifecycle to react to design changes with minimal negative impacts on time and cost of manufacturing. SM often relates to Internet of Things (IoT) and Cyber-Physical Systems (CPS) in the way that hardware and software facilitate the communication of actions in different parts of a manufacturing system. A common method of communication among different parts of a SM system is the Internet. Sensors, platforms, and services that communicate in SM need to securely connect to the Internet and communicate with one another in a standard language. An increasingly popular language for communication in IoT in general and more specifically for hardware in manufacturing is MTConnect. The goal of this work is to demonstrate an approach for development of a platform that collects high frequency data from MTConnect and non-MTConnect platforms, enables machines and platforms to directly communicate with one another, processes the collected data, and securely communicates with Internet and cloud without the need of a static IP address. More specifically, the proposed platform consists of two separate sections, a Local Area Network (LAN) and an Internet Area Network (IAN). The LAN communicates with machines via MTConnect, performs fog computing, and transfers the data to the IAN. The IAN section receives the data from LAN while acquiring high frequency data from sensors. This platform then communicates with Internet-connected devices and web APIs for different tasks such as inserting data to a database, providing data for web apps or smartphone apps, sending alerting messages, and communicating with cloud services such as Google Clouds, Amazon Web Services, If-This-Then-That (IFTTT), and Particle Clouds.