Graduate Exam Abstract
November 29, 2017, 10:00 am - 12:00 pm
ERC Electronic Classroom at Foothills Campus
A Foray into Neural Network Modeling and Control of Particle Accelerators
Abstract: Particle accelerators are host to myriad nonlinear and complex physical phenomena. They often involve a multitude of interacting systems, are subject to tight performance demands, and should be able to run for extended periods of time with minimal interruptions. Often times, traditional control techniques cannot fully meet these requirements, and accelerator facilities rely heavily on highly skilled human operators to control the accelerator. One promising but relatively unexplored avenue to improve both operational performance and efficiency is to develop machine learning based approaches to address these challenges. Machine learning constitutes a versatile set of techniques that are particularly well-suited to modeling, control, and diagnostic analysis of complex, time-varying systems, as well as systems with large parameter spaces. These techniques can be used in conjunction with actual system data, thereby accounting for noise, variable delays, subtle statistical correlations, and complex effects that may not be addressed in models based on first-principles alone. Because of the way these techniques interact with the data, controllers based on machine learning are able to account for physical characteristics of systems which (a) have many interactions between a large number of parameters, (b) are not able to be realistically or completely modeled through analytic or standard physics-based simulation methods (due to practical or theoretical limitations), or (c) vary significantly over time or involve behavior over multiple timescales. Within machine learning, neural networks are especially appealing tools for modeling and control applications due to their functional flexibility and subsequent ability to operate effectively for many different kinds of tasks. Taking these factors into account, the use of neural network-based modeling and control techniques could be of significant benefit to a wide range of particle accelerator control tasks. Particle accelerators are also ideal test-beds for these techniques, given their complex behavior and the large amount of associated measured data available. Many early attempts to apply neural networks to particle accelerators (mostly in the early 1990s) obtained mixed results due to the relative immaturity of the technology at that time. The central aim of this thesis is to explore this avenue for advancing the state of the art in modeling and control of particle accelerators. Here, some of the challenges of particle accelerator control are described, recent advances in neural network techniques for modeling and control are highlighted, some promising avenues for incorporating neural networks into particle accelerator control systems are discussed, and several experimental and simulation-based studies in neural network modeling and control for particle accelerators are described. These studies have provided initial results for a variety of tasks that are relevant to particle accelerators and have helped pave the way for further development. Further development of these approaches could increase the time-and energy-efficiency of these machines, increase automation, improve beam quality, and act as enabling technology for some aspects of future particle accelerators and their applications.
Adviser: Sandra Biedron
Co-Adviser: Stephen Milton
Non-ECE Member: Thomas Johnson, Environmental and Radiological Health Sciences
Member 3: Edwin Chong, Electrical and Computer Engineering, Mathematics
Addional Members: NA
- Edelen, A.L., Biedron, S.G., Chase, B.E., Edstrom, D., Milton, S.V., Stabile, P., “Neural Networks for Modeling and Control of Particle Accelerators,” IEEE Transactions on Nuclear Science, vol. 63, no. 2, pp. 878-897, Apr. 2016. (invited paper)
- Edelen, J.P., Edelen, A.L., Bowring, D., Chase, B.E., Steimel, J., Biedron, S.G., Milton, S.V., “First Principles Modeling of RFQ Cooling System and Resonant Frequency Responses for Fermilab's PIP-II Injector Test,” IEEE Transactions on Nuclear Science, vol. PP, no. 99, pp. 1, Dec. 2016.
- Hall, C.C., Biedron, S.G., Edelen, A.L., Milton, S.V., Benson, S., Douglas, D., Li, R., Tennant, C.D., Carlsten, B.E., “Measurement and simulation of the impact of coherent synchrotron radiation on the Jefferson Laboratory energy recovery linac electron beam,” Phys. Rev. ST Accel. Beams, vol. 18, no. 3, 2015.
- Edelen, A.L., Edelen, J.P., Biedron, S.G., Milton, S.V., van der Slot, P., Using a Neural Network Control Policy for Rapid Switching Between Beam Parameters in an FEL, Proc. of the 2017 Free Electron Laser Conference, Santa Fe, NM, Aug. 20 – 25, 2017.
- Edelen, A.L., Biedron, S.G., Milton, S.V., Bowring, D., Chase, B.E., Edelen, J.P., Steimel, J., Resonant Frequency Control for the PIP-II Injector Test RFQ: Control Framework and Initial Results, Proc. of the 2016 North American Particle Accelerator Conference, Chicago, IL, Oct. 9 – 11, 2016.
- Edelen, A.L., Edelen, J.P., Biedron, S.G., Milton, S.V., First Steps Toward Incorporating Image-Based Diagnostics Directly into Particle Accelerator Control Systems Using Convolutional Neural Networks, Proc. of the 2016 North American Particle Accelerator Conference, Chicago, IL, Oct. 9-11, 2016
- Steimel, J., Baffes, C., Berrutti, P., Cariero, J.-P., Edelen, A., Edelen, J., Hoff, M., Khabiboulline, T., Lambert, A., Li, D., Luo, T., Prost, L., Scarpine, V., Shemyakin, A., Lalitha Sista, V., Staples, J., Virostek, S., Beam Commissioning Status and Results of the FNAL PIP2IT Linear Accelerator RFQ, Proc. of the 28th Linear Accelerator Conference (LINAC ’16), East Lansing, MI, Sep. 25 – 30 , 2016.
- Baffes, C., Alvarez, M., Andrews, R., Chen, A., Czajkowski, J., Derwent, P., Edelen, J., Hanna, B., Hartsell, B., Kendziora, K., Mitchell, D., Prost, L., Scarpine, V., Shemyakin, A., Steimel, J., Zuchnik, T., Edelen, A., Installation Progress at the PIP-II Injector Test at Fermilab, Proc. of the 2016 North American Particle Accelerator Conference, Chicago, IL, Oct. 9-11, 2016.
- Edelen, A.L., Biedron, S.G., Milton, S.V., Bowring, D., Chase, B.E., Edelen, J.P., Steimel, J., Neural network model of the PXIE RFQ cooling system and resonant frequency response, Proc. of the 2016 International Particle Accelerator Conference, Busan, Korea, May 8 – 13, 2016.
- Bowring, D., Chase, B., Czajkowski, J., Edelen, J, Nicklaus, D., Steimel, J., Zuchnik, T., Edelen, A., Biedron, S., Milton, S., Resonance control for Fermilab's PXIE RFQ, Proc. of the 2016 International Particle Accelerator Conference, Busan, Korea, May 8 – 13, 2016.
- Edelen, A.L., Biedron, S.G., Chase, B.E., Crawford, D.E., Eddy, N., Edstrom Jr., D., Harms, E.R., Ruan, J., Santucci, J.K., Stabile, P., Initial Experimental Results of a Machine Learning-Based Temperature Control System for an RF Gun. Proc. of the 2015 International Particle Accelerator Conference, Richmond, VA, May 3 – 8, 2015.
- Hall, C., Carlsten, B.E., Biedron, S., Edelen, A., Milton, S.V., Benson, S., Douglas, D., Li, R., Tennant, C., Coherent Synchrotron Radiation in Energy Recovery Linacs, Proc. of the 2015 International Particle Accelerator Conference, Richmond, VA, May 3 – 8, 2015.
- Biedron, S.G., Milton, S.V., D’Audney, A., Edelen, J., Einstein, J. Harris, J., Hall, C., Horovits, K., Martinez, J., Morin, A., Sipahi, N., Sipahi, T., Williams, J., The CSU Accelerator and FEL Facility, Proc. of the APS April Meeting 2014, Savannah, Georgia, Apr. 5 – 8, 2014.
- Morin, A., Biedron, S.G., Benson, S., Douglas, D., Milton, S.V., Tennant, C., Control Systems Development for the Thomas Jefferson National Accelerator Facility Free Electron Laser & Energy Recovery Linac: Analysis of Trajectory Response Data. Poster presented at the 14th International Conference on Accelerator & Large Experimental Physics Control Systems, San Francisco, CA, Oct. 7 – 11, 2013.
Program of Study: