More Like this

1. Manufacturing of Soft Magnets using the Fused Filament Fabrication Process

   https://doi.org/doi.org/10.13140/RG.2.2.32731.90407  

   Hasanov, Seymur ; Fidan, Ismail ( January 2020 , MatEDU Resource Center)

   Stoebe, Thomas (Ed.)

   Rare-earth (RE) materials are currently used to fabricate permanent magnets through various additive manufacturing (AM) methods. Fused filament fabrication (FFF) is one of the most commonly used polymer-based AM methods and has recently been used to produce metal-matrix composites, known as “green parts,” using a metal powder-infused filament. The FFF method has gained much attention in various industries including the automotive, aerospace, and medical fields. Therefore, involving RE in the FFF process using magnetic powder-infused filaments promises to result in the fabrication of low-cost, efficient, and complex magnetic components based on application areas. This module introduces the FFF process and provides a case study for high school and technical college students to gain a fundamental understanding of how magnetic powders are infused and how parts are fabricated using this method. 

   more » « less
2. Creating and Assessing an Upper Division Additive Manufacturing Course and Laboratory to Enhance Undergraduate Research and Innovation

   https://doi.org/10.18260/1-2--30124  

   Maloney, Patricia A ; Li, Bingbing ; Zhang, Meng ; Cong, Weilong ( June 2018 , 2018 ASEE Annual Conference & Exposition)

   This NSF IUSE project is on the Exploration and Design Tier and the Engaged Student Learning Track. It is aimed at better preparing the country’s professional workforce in the renaissance of U.S. skilled manufacturing by creating new personnel proficient in additive manufacturing (AM). AM is mainstream; it has the potential to bring jobs back to the U.S. and add to the nation’s global competitiveness. AM is the process of joining materials to make objects from 3D data in a layer upon layer fashion. The objectives are to develop, assess, revise, and disseminate an upper division course and laboratory, “Additive Manufacturing,” and to advance undergraduate and K-12 student research and creative inquiry activities as well as faculty expertise at three diverse participating universities: Texas Tech, California State-Northridge, and Kansas State. This research/pedagogical team contains a mechanical engineering professor at each university to develop and teach the course, as well as a sociologist trained in K-12 outreach, course assessment, and human subjects research to accurately determine the effects on K-12 and undergraduate students. The proposed course will cover extrusion-based, liquid-based, and powder-based AM processes. For each technology, fundamentals, applications, and advances will be discussed. Students will learn solutions to AM of polymers, metals, and ceramics. Two lab projects will be built to provide hands-on experiences on a variety of state-of-the-art 3D printers. To stimulate innovation, students will design, fabricate, and measure test parts, and will perform experiments to explore process limits and tackle real world problems. They will also engage K-12 students through video demonstrations and mentorship, thus developing presentation skills. Through the project, different pedagogical techniques and assessment tools will be utilized to assess and improve engineering education at both the undergraduate and K-12 levels through varied techniques: i) undergraduate module lesson plans that are scalable to K-12 levels, ii) short informational video lessons created by undergraduates for K-12 students with accompanying in-person mentorship activities at local high schools and MakerSpaces, iii) pre- and post-test assessments of undergraduates’ and K-12 participating students’ AM knowledge, skills, and perceptions of self-efficacy, and iv) focus groups to learn about student concerns/learning challenges. We will also track students institutionally and into their early careers to learn about their use of AM technology professionally. 

   more » « less
3. Creating and Assessing an Upper Division Additive Manufacturing Course and Laboratory to Enhance Undergraduate Research and Innovation

   Maloney, P ; Li, B ; Zhang, M ; Cong, W. ( June 2018 , ASEE annual conference & exposition)

   This NSF IUSE project is on the Exploration and Design Tier and the Engaged Student Learning Track. It is aimed at better preparing the country’s professional workforce in the renaissance of U.S. skilled manufacturing by creating new personnel proficient in additive manufacturing (AM). AM is mainstream; it has the potential to bring jobs back to the U.S. and add to the nation’s global competitiveness. AM is the process of joining materials to make objects from 3D data in a layer upon layer fashion. The objectives are to develop, assess, revise, and disseminate an upper division course and laboratory, “Additive Manufacturing,” and to advance undergraduate and K-12 student research and creative inquiry activities as well as faculty expertise at three diverse participating universities: Texas Tech, California State Northridge, and Kansas State. This research/pedagogical team contains a mechanical engineering professor at each university to develop and teach the course, as well as a sociologist trained in K-12 outreach, course assessment, and human subjects research to accurately determine the effects on K-12 and undergraduate students. The proposed course will cover extrusion-based, liquid-based, and powder-based AM processes. For each technology, fundamentals, applications, and advances will be discussed. Students will learn solutions to AM of polymers, metals, and ceramics. Two lab projects will be built to provide hands-on experiences on a variety of state-of-the-art 3D printers. To stimulate innovation, students will design, fabricate, and measure test parts, and will perform experiments to explore process limits and tackle real world problems. They will also engage K-12 students through video demonstrations and mentorship, thus developing presentation skills. Through the project, different pedagogical techniques and assessment tools will be utilized to assess and improve engineering education at both the undergraduate and K-12 levels through varied techniques: i) undergraduate module lesson plans that are scalable to K-12 levels, ii) short informational video lessons created by undergraduates for K-12 students with accompanying in-person mentorship activities at local high schools and MakerSpaces, iii) pre- and post-test assessments of undergraduates’ and K-12 participating students’ AM knowledge, skills, and perceptions of self-efficacy, and iv) focus groups to learn about student concerns/learning challenges. We will also track students institutionally and into their early careers to learn about their use of AM technology professionally. 

   more » « less
4. Development of a Multiscale Experimentation and Visualization Module for Undergraduate Mechanics Education

   https://doi.org/10.18260/1-2--34450  

   Herren, Blake ; Mason, Nyree ; Akasheh, Firas ; Okudan-Kremer, Gül ; Siddique, Zahed ; Liu, Yingtao ( June 2020 , 2020 ASEE Virtual Annual Conference)

   null (Ed.)

   Classical mechanics courses are taught to most engineering disciplinary undergraduate students. Due to the recent advancements of multiscale analysis and practice, necessary reforms need to be investigated and explored for classical mechanics courses to address the materials’ mechanics behaviors across multiple length scales. This enhanced understanding is needed for engineering students to consider materials more broadly. This paper presents a recent effort for the development of a multiscale materials and mechanics experimentation (M3E) module that can be potentially implemented in undergraduate mechanics courses, including Statics, Dynamics, Strength of Materials, and Design of Mechanical (Machine) Components. The developed education module introduces the concepts of multiscale materials behavior and microstructures in the form of micro and macro-scales. At the micro-scale, both 3D printed aluminum and cold-rolled aluminum samples were characterized using scanning electron microscope. Microstructures, including grains, grain boundaries, dislocation, precipitates, and micro-voids, were demonstrated to students. At the macro-scale, experiments following ASTM standards were conducted and full strain fields carried by all the samples were analyzed using digital image correlation method. The experimental data were organized and presented to the students in the developed M3E module. The implementation of the developed module in undergraduate mechanics classes allows students to not only visualize materials behavior under various load conditions, but also understand the reasons behind classical mechanics properties. To assess the effectiveness of the developed M3E education module, an evaluation question was developed. Students are required to classify key mechanics, materials, and processing concepts at both micro and macroscales. More than 40 fundamental concepts and keywords are included in the tests. The study outcomes and effectiveness of the M3E education module will be reported in this paper. 

   more » « less
5. A Deep Learning-Based Real-time Seizure Detection System

   https://doi.org/10.1109/SPMB50085.2020.9353623  

   Shawki, N. ; Elseify, T. ; Cap, T. ; Shah, V. ; Obeid, I. ; Picone, J. ( December 2020 , Proceedings of the IEEE Signal Processing in Medicine and Biology Symposium (SPMB))

   Obeid, Iyad Selesnick (Ed.)

   Electroencephalography (EEG) is a popular clinical monitoring tool used for diagnosing brain-related disorders such as epilepsy [1]. As monitoring EEGs in a critical-care setting is an expensive and tedious task, there is a great interest in developing real-time EEG monitoring tools to improve patient care quality and efficiency [2]. However, clinicians require automatic seizure detection tools that provide decisions with at least 75% sensitivity and less than 1 false alarm (FA) per 24 hours [3]. Some commercial tools recently claim to reach such performance levels, including the Olympic Brainz Monitor [4] and Persyst 14 [5]. In this abstract, we describe our efforts to transform a high-performance offline seizure detection system [3] into a low latency real-time or online seizure detection system. An overview of the system is shown in Figure 1. The main difference between an online versus offline system is that an online system should always be causal and has minimum latency which is often defined by domain experts. The offline system, shown in Figure 2, uses two phases of deep learning models with postprocessing [3]. The channel-based long short term memory (LSTM) model (Phase 1 or P1) processes linear frequency cepstral coefficients (LFCC) [6] features from each EEG channel separately. We use the hypotheses generated by the P1 model and create additional features that carry information about the detected events and their confidence. The P2 model uses these additional features and the LFCC features to learn the temporal and spatial aspects of the EEG signals using a hybrid convolutional neural network (CNN) and LSTM model. Finally, Phase 3 aggregates the results from both P1 and P2 before applying a final postprocessing step. The online system implements Phase 1 by taking advantage of the Linux piping mechanism, multithreading techniques, and multi-core processors. To convert Phase 1 into an online system, we divide the system into five major modules: signal preprocessor, feature extractor, event decoder, postprocessor, and visualizer. The system reads 0.1-second frames from each EEG channel and sends them to the feature extractor and the visualizer. The feature extractor generates LFCC features in real time from the streaming EEG signal. Next, the system computes seizure and background probabilities using a channel-based LSTM model and applies a postprocessor to aggregate the detected events across channels. The system then displays the EEG signal and the decisions simultaneously using a visualization module. The online system uses C++, Python, TensorFlow, and PyQtGraph in its implementation. The online system accepts streamed EEG data sampled at 250 Hz as input. The system begins processing the EEG signal by applying a TCP montage [8]. Depending on the type of the montage, the EEG signal can have either 22 or 20 channels. To enable the online operation, we send 0.1-second (25 samples) length frames from each channel of the streamed EEG signal to the feature extractor and the visualizer. Feature extraction is performed sequentially on each channel. The signal preprocessor writes the sample frames into two streams to facilitate these modules. In the first stream, the feature extractor receives the signals using stdin. In parallel, as a second stream, the visualizer shares a user-defined file with the signal preprocessor. This user-defined file holds raw signal information as a buffer for the visualizer. The signal preprocessor writes into the file while the visualizer reads from it. Reading and writing into the same file poses a challenge. The visualizer can start reading while the signal preprocessor is writing into it. To resolve this issue, we utilize a file locking mechanism in the signal preprocessor and visualizer. Each of the processes temporarily locks the file, performs its operation, releases the lock, and tries to obtain the lock after a waiting period. The file locking mechanism ensures that only one process can access the file by prohibiting other processes from reading or writing while one process is modifying the file [9]. The feature extractor uses circular buffers to save 0.3 seconds or 75 samples from each channel for extracting 0.2-second or 50-sample long center-aligned windows. The module generates 8 absolute LFCC features where the zeroth cepstral coefficient is replaced by a temporal domain energy term. For extracting the rest of the features, three pipelines are used. The differential energy feature is calculated in a 0.9-second absolute feature window with a frame size of 0.1 seconds. The difference between the maximum and minimum temporal energy terms is calculated in this range. Then, the first derivative or the delta features are calculated using another 0.9-second window. Finally, the second derivative or delta-delta features are calculated using a 0.3-second window [6]. The differential energy for the delta-delta features is not included. In total, we extract 26 features from the raw sample windows which add 1.1 seconds of delay to the system. We used the Temple University Hospital Seizure Database (TUSZ) v1.2.1 for developing the online system [10]. The statistics for this dataset are shown in Table 1. A channel-based LSTM model was trained using the features derived from the train set using the online feature extractor module. A window-based normalization technique was applied to those features. In the offline model, we scale features by normalizing using the maximum absolute value of a channel [11] before applying a sliding window approach. Since the online system has access to a limited amount of data, we normalize based on the observed window. The model uses the feature vectors with a frame size of 1 second and a window size of 7 seconds. We evaluated the model using the offline P1 postprocessor to determine the efficacy of the delayed features and the window-based normalization technique. As shown by the results of experiments 1 and 4 in Table 2, these changes give us a comparable performance to the offline model. The online event decoder module utilizes this trained model for computing probabilities for the seizure and background classes. These posteriors are then postprocessed to remove spurious detections. The online postprocessor receives and saves 8 seconds of class posteriors in a buffer for further processing. It applies multiple heuristic filters (e.g., probability threshold) to make an overall decision by combining events across the channels. These filters evaluate the average confidence, the duration of a seizure, and the channels where the seizures were observed. The postprocessor delivers the label and confidence to the visualizer. The visualizer starts to display the signal as soon as it gets access to the signal file, as shown in Figure 1 using the “Signal File” and “Visualizer” blocks. Once the visualizer receives the label and confidence for the latest epoch from the postprocessor, it overlays the decision and color codes that epoch. The visualizer uses red for seizure with the label SEIZ and green for the background class with the label BCKG. Once the streaming finishes, the system saves three files: a signal file in which the sample frames are saved in the order they were streamed, a time segmented event (TSE) file with the overall decisions and confidences, and a hypotheses (HYP) file that saves the label and confidence for each epoch. The user can plot the signal and decisions using the signal and HYP files with only the visualizer by enabling appropriate options. For comparing the performance of different stages of development, we used the test set of TUSZ v1.2.1 database. It contains 1015 EEG records of varying duration. The any-overlap performance [12] of the overall system shown in Figure 2 is 40.29% sensitivity with 5.77 FAs per 24 hours. For comparison, the previous state-of-the-art model developed on this database performed at 30.71% sensitivity with 6.77 FAs per 24 hours [3]. The individual performances of the deep learning phases are as follows: Phase 1’s (P1) performance is 39.46% sensitivity and 11.62 FAs per 24 hours, and Phase 2 detects seizures with 41.16% sensitivity and 11.69 FAs per 24 hours. We trained an LSTM model with the delayed features and the window-based normalization technique for developing the online system. Using the offline decoder and postprocessor, the model performed at 36.23% sensitivity with 9.52 FAs per 24 hours. The trained model was then evaluated with the online modules. The current performance of the overall online system is 45.80% sensitivity with 28.14 FAs per 24 hours. Table 2 summarizes the performances of these systems. The performance of the online system deviates from the offline P1 model because the online postprocessor fails to combine the events as the seizure probability fluctuates during an event. The modules in the online system add a total of 11.1 seconds of delay for processing each second of the data, as shown in Figure 3. In practice, we also count the time for loading the model and starting the visualizer block. When we consider these facts, the system consumes 15 seconds to display the first hypothesis. The system detects seizure onsets with an average latency of 15 seconds. Implementing an automatic seizure detection model in real time is not trivial. We used a variety of techniques such as the file locking mechanism, multithreading, circular buffers, real-time event decoding, and signal-decision plotting to realize the system. A video demonstrating the system is available at: https://www.isip.piconepress.com/projects/nsf_pfi_tt/resources/videos/realtime_eeg_analysis/v2.5.1/video_2.5.1.mp4. The final conference submission will include a more detailed analysis of the online performance of each module. ACKNOWLEDGMENTS Research reported in this publication was most recently supported by the National Science Foundation Partnership for Innovation award number IIP-1827565 and the Pennsylvania Commonwealth Universal Research Enhancement Program (PA CURE). Any opinions, findings, and conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the official views of any of these organizations. REFERENCES [1] A. Craik, Y. He, and J. L. Contreras-Vidal, “Deep learning for electroencephalogram (EEG) classification tasks: a review,” J. Neural Eng., vol. 16, no. 3, p. 031001, 2019. https://doi.org/10.1088/1741-2552/ab0ab5. [2] A. C. Bridi, T. Q. Louro, and R. C. L. Da Silva, “Clinical Alarms in intensive care: implications of alarm fatigue for the safety of patients,” Rev. Lat. Am. Enfermagem, vol. 22, no. 6, p. 1034, 2014. https://doi.org/10.1590/0104-1169.3488.2513. [3] M. Golmohammadi, V. Shah, I. Obeid, and J. Picone, “Deep Learning Approaches for Automatic Seizure Detection from Scalp Electroencephalograms,” in Signal Processing in Medicine and Biology: Emerging Trends in Research and Applications, 1st ed., I. Obeid, I. Selesnick, and J. Picone, Eds. New York, New York, USA: Springer, 2020, pp. 233–274. https://doi.org/10.1007/978-3-030-36844-9_8. [4] “CFM Olympic Brainz Monitor.” [Online]. Available: https://newborncare.natus.com/products-services/newborn-care-products/newborn-brain-injury/cfm-olympic-brainz-monitor. [Accessed: 17-Jul-2020]. [5] M. L. Scheuer, S. B. Wilson, A. Antony, G. Ghearing, A. Urban, and A. I. Bagic, “Seizure Detection: Interreader Agreement and Detection Algorithm Assessments Using a Large Dataset,” J. Clin. Neurophysiol., 2020. https://doi.org/10.1097/WNP.0000000000000709. [6] A. Harati, M. Golmohammadi, S. Lopez, I. Obeid, and J. Picone, “Improved EEG Event Classification Using Differential Energy,” in Proceedings of the IEEE Signal Processing in Medicine and Biology Symposium, 2015, pp. 1–4. https://doi.org/10.1109/SPMB.2015.7405421. [7] V. Shah, C. Campbell, I. Obeid, and J. Picone, “Improved Spatio-Temporal Modeling in Automated Seizure Detection using Channel-Dependent Posteriors,” Neurocomputing, 2021. [8] W. Tatum, A. Husain, S. Benbadis, and P. Kaplan, Handbook of EEG Interpretation. New York City, New York, USA: Demos Medical Publishing, 2007. [9] D. P. Bovet and C. Marco, Understanding the Linux Kernel, 3rd ed. O’Reilly Media, Inc., 2005. https://www.oreilly.com/library/view/understanding-the-linux/0596005652/. [10] V. Shah et al., “The Temple University Hospital Seizure Detection Corpus,” Front. Neuroinform., vol. 12, pp. 1–6, 2018. https://doi.org/10.3389/fninf.2018.00083. [11] F. Pedregosa et al., “Scikit-learn: Machine Learning in Python,” J. Mach. Learn. Res., vol. 12, pp. 2825–2830, 2011. https://dl.acm.org/doi/10.5555/1953048.2078195. [12] J. Gotman, D. Flanagan, J. Zhang, and B. Rosenblatt, “Automatic seizure detection in the newborn: Methods and initial evaluation,” Electroencephalogr. Clin. Neurophysiol., vol. 103, no. 3, pp. 356–362, 1997. https://doi.org/10.1016/S0013-4694(97)00003-9. 

   more » « less