Figure 13 - uploaded by Stephen Milton
Content may be subject to copyright.
The basic elements of a model predictive control scheme. Np is the prediction horizon, Nm is the number of previous measured values used for modeling, k is the present time step, Nc is the control horizon, ucv are the controlled variables, um are measured variables, and yp is the predicted plant output.
Source publication
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. One p...
Similar publications
An adaptive control law has been designed for robust attitude tracking of a spacecraft with uncertain inertia matrix, torque coupling coefficients and exogenous disturbances. The strict block feedback structure of the attitude dynamics is exploited in designing a back stepping control law, which depends on a compensator variable whose dynamics is d...
The design of controllers and observers often relies on first order models of the system in question. These models are often obtained either through step-response tests, through on-line or off-line identification, or through developing a mathematical model. When the system in question has unknown or uncertain parameters, the developed model also co...
The quadrotor control has been one of the benchmark control problems. It is considered as an under-actuated, multivariable and high nonlinear system due to its dynamics, having strong coupling between translation and angular motion and affected by external disturbances associated with flight environment. Therefore, there is a need to design a robus...
Citations
... Additionally, ML can be employed to analyze and interpret exper-imental data more effectively, allowing for faster and more precise extraction of physical quantities such as cross-sections, particle trajectories, and event classification. These advancements are helping to bridge the gap between traditional simulation methods and the vast amounts of data generated by modern high-energy physics experiments, making it possible to explore new regions of parameter space and extract insights more efficiently [74][75][76][77][78][79][80][81][82][83][84]. This collaboration between the simulation and the ML is paving the way for deeper insights into the hadron physics. ...
Generative Adversarial Networks (GANs) are influential machine learning models that have gained prominence across various research fields, including high-energy physics (HEP) simulations. Among them, Conditional GANs (CGANs) offer a unique advantage by conditioning on specific parameters, making them particularly well-suited for meson studies. Given the limited size of the available meson datasetbased on the quark content, quantum numbers, mass, and widththis study, for the first time, applies the CGAN framework to augment the dataset of mesons, while preserving the inherent characteristics of the original data. Using this enhanced dataset, we employ the CGAN model to estimate the mass and width of both ordinary and exotic mesons, based solely on their quark content and quantum numbers. Our CGANbased predictions highlight the framework's potential as a reliable tool for meson property estimation, providing valuable insights for future research in particle physics.
... Traditional methods, such as physics-based approaches or linear models, often cannot handle complex non-linear relationships. On the other hand, one of the latest technologies in artificial intelligence, such as deep neural networks (DNN) [10], has proven effective in learning complex data patterns, as several studies have shown [11], [12], [13], [14]. DNN can learn from past data to make future predictions with a higher degree of accuracy, especially in situations where the relationship patterns between variables are non-linear [15]. ...
The consistency of cyclotron ion source output is crucial for ensuring high-quality radioactive isotope production, such as fluorine-18, which is widely used in positron emission tomography (PET) for cancer diagnosis. This study aims to develop a predictive model for cyclotron ion source output using deep neural networks (DNN). The model was trained on historical operational data from the Cyclotron at Dharmais Cancer Hospital, including variables such as ion source voltage, ion source current, bias voltage, bias current, gas flow, high vacuum, and magnetism. The DNN effectively captured the complex non-linear relationships among these parameters, overcoming the limitations of traditional physical models and linear methods. Evaluation results showed great performance, with a test MSE of 0.00178 and an MAE of 0.03064, demonstrating the model's ability to provide accurate predictions. The findings highlight the potential of DNN for improving cyclotron efficiency and reliability, contributing to better outcomes in nuclear medicine services.
... The real-time machine learning renders enormous potential for qualitative or quantitative character screening with speed breeding techniques in a controlled environment ecosystem. The study may be necessary to investigate speed breeding techniques intervention in biometric, biofortification, nutrient or water dynamics screening, tissue cultured plant regeneration examination, and physiochemical characterization with artificial intelligence in millets under controlled growing conditions (Edelen et al. 2016;Kundu et al. 2021). ...
Millet breeding focuses on improving essential traits such as grain yield, head structure, tiller production, early maturity, reduced plant height, biomass, digestibility and key nutrients like vitamin B1, lysine, cysteine and methionine. Traditional breeding, especially in open environments, can take 9–17 years to release a new variety, whereas speed breeding in controlled environments shortens this to 5–9 years. This accelerated process tackles challenges like male sterility, self-incompatibility, seed shattering, inbreeding depression and embryo abortion. Techniques such as rapid single-seed descent enable the creation of near-homozygous lines in 1–2 years, allowing finger millet to achieve up to five generations per year. Indoor phenotyping platforms enhance speed breeding by providing detailed, consistent monitoring of plant traits. High-throughput systems in controlled settings like growth chambers or glasshouses allow for non-invasive assessment of entire crop canopies, measuring traits such as leaf expansion, width, phyllochron and stomatal conductance. This precise phenotyping accelerates trait evaluation and selection, facilitating the development of superior millet varieties. Supported by advanced phenotyping and gene-editing tools, speed breeding offers a robust solution for improving key agronomic traits, significantly reducing breeding time in controlled environments
... Machine learning (ML) has been identified as having the potential for significant impact on the modeling, operation, and control of particle accelerators [1,2]. For machine diagnostics specifically, there have been numerous efforts to improve measurement capabilities and detect faulty instruments. ...
Industrial particle accelerators typically operate in dirtier environments than research accelerators, leading to increased noise in RF and electronic systems. Furthermore, given that industrial accelerators are mass produced, less attention is given to optimizing the performance of individual systems. As a result, industrial accelerators tend to underperform their own hardware capabilities. Improving signal processing for these machines will improve cost and time margins for deployment, helping to meet the growing demand for accelerators for medical sterilization, food irradiation, cancer treatment, and imaging. Our work focuses on using machine learning techniques to reduce noise in RF signals used for pulse-to-pulse feedback in industrial accelerators. Here we review our algorithms and observed results for simulated RF systems, and discuss next steps with the ultimate goal of deployment on industrial systems.
... Recently, improved techniques from the fields of machine learning (ML) and artificial intelligence (AI) have been incorporated into the design of control systems for particle accelerators. In particular, techniques based on neural networks (NNs) are well-suited to modeling, control, and diagnostic analysis of complex, time-varying systems, and systems with large parameter spaces [11,12]. ...
Undulators are used in storage rings to produce extremely brilliant synchrotron radiation. In the ideal case, a perfectly tuned undulator always has a first and second field integrals equal to zero. But, in practice, field integral changes during gap movements can never be avoided for real-life devices. As they significantly impact the circulating electron beam, there is the need to routinely compensate such effects. Deep Neural Networks can be used to predict the distortion in the closed orbit induced by the undulator gap variations on the circulating electron beam. In this contribution several current state-of-the-art deep learning algorithms were trained on measurements from PETRA~III. The different architecture performances are then compared to identify the best model for the gap-induced distortion compensation.
... This information is valuable for control systems to respond appropriately, activating backup systems or rerouting processes as needed. 6. Adaptive control strategies: AI-driven adaptive control strategies can adjust the control algorithms in realtime based on changing diagnostic conditions defined on complex control models [37]. This adaptability is particularly useful in environments where operational parameters may fluctuate. ...
... 9. Continuous model learning: ML models can be designed for continuous learning, adapting to evolving system dynamics over time. From well-known neutral networks technics [37], to latest advances in reinforcement learning [39], this adaptability is crucial in environments where the characteristics of instruments or processes may change like in an experimental plant. 10. ...
As an integral part of the European strategy for advancing fusion-generated electricity, IFMIF-DONES represents a high-intensity neutron irradiation plant with the main purpose of assessing the suitability of materials for fusion reactor applications. Its primary mission is to examine how materials respond to irradiation within a neutron flux that mimics the conditions expected in the first wall of the proposed DEMO reactor, which is intended to succeed ITER. Consequently, IFMIF-DONES, whose construction is slated to commence shortly, plays a pivotal role in aiding the development, approval, and safe operation of DEMO, as well as future fusion power plants. This paper provides a quick overview of the current development of the IFMIF-DONES neutron source with a particular snapshot of the present engineering design status for what concerns the instrumentation and control systems together with its complex diagnostics, that guarantees the safe monitoring, supervision and regulation of all operations. The current status of design, after the completion of the preliminary design phase is presented, as well as the existing and future plans for their integration also using some of the new capabilities offered by Artificial Intelligence tools.
... The use of ML for particle accelerator applications has grown in recent years including, but not limited to, diagnostics [15][16][17][18][19][20][21], anomaly detection/forecasting/classification [3,4,[22][23][24], and optimization/controls [25][26][27][28][29][30][31][32]. ...
Particle accelerators are complex and comprise thousands of components, with many pieces of equipment running at their peak power. Consequently, they can fault and abort operations for numerous reasons, lowering efficiency and science output. To avoid these faults, we apply anomaly detection techniques to predict unusual behavior and perform preemptive actions to improve the total availability. Supervised Machine Learning (ML) techniques such as Siamese Neural Network (SNN) models can outperform the often-used unsupervised or semi-supervised approaches for anomaly detection by leveraging the label information. One of the challenges specific to anomaly detection for particle accelerators is the data’s variability due to accelerator configuration changes within a production run of several months. ML models fail at providing accurate predictions when data changes due to changes in the configuration. To address this challenge, we include the configuration settings into our models and training to improve the results. Beam configurations are used as a conditional input for the model to learn any cross-correlation between the data from different conditions and retain its performance. We employ Conditional Siamese Neural Network (CSNN) models and Conditional Variational Auto Encoder (CVAE) models to predict errant beam pulses at the Spallation Neutron Source (SNS) under different system configurations and compare their performance. We demonstrate that CSNNs outperform CVAEs in our application.
... To overcome these challenges, we develop a virtual diagnostics model based on artificial neural networks (ANNs) and shot-to-shot measurement data of both electron and X-ray beam parameters. ANNs are powerful tools for modeling complex nonlinear relationships, and exploration of their utility to overcome the limitations of conventional methods for accelerator optimization, tuning, and modeling is underway [14][15][16][17] . The majority of machine learning models for XFELs have primarily focused only on the electron beam for tasks such as accelerator and undulator tuning and optimization 18,19 , with one study incorporating X-ray spectrometer data 20 . ...
Knowledge of x-ray free electron lasers’ (XFELs) pulse characteristics delivered to a sample is crucial for ensuring high-quality x-rays for scientific experiments. XFELs’ self-amplified spontaneous emission process causes spatial and spectral variations in x-ray pulses entering a sample, which leads to measurement uncertainties for experiments relying on multiple XFEL pulses. Accurate in-situ measurements of x-ray wavefront and energy spectrum incident upon a sample poses challenges. Here we address this by developing a virtual diagnostics framework using an artificial neural network (ANN) to predict x-ray photon beam properties from electron beam properties. We recorded XFEL electron parameters while adjusting the accelerator’s configurations and measured the resulting x-ray wavefront and energy spectrum shot-to-shot. Training the ANN with this data enables effective prediction of single-shot or average x-ray beam output based on XFEL undulator and electron parameters. This demonstrates the potential of utilizing ANNs for virtual diagnostics linking XFEL electron and photon beam properties.
... In recent years, artificial-intelligence-based methods have been introduced in the literature for solving various complex problems [24]. The applications of artificial neural networks (ANNs) can be seen in circuit and electromagnetic theory [25,26], the fuel ignition model [27], motor induction models [28], the Thomas-Fermi model [29], doubly singular nonlinear systems [28], nanotechnology [30], nanofluidics [31], nonlinear prey-predator models [32], nonlinear equations [33], Troesch's problem [34], optimal control [35], signal processing [36], and the modeling and control of particle accelerators [37]. Pinsky and Datye [38] used the finite element method for the analysis of the incised cornea of the human eye. ...
In this paper, a hybrid cuckoo search technique is combined with a single-layer neural network (BHCS-ANN) to approximate the solution to a differential equation describing the curvature shape of the cornea of the human eye. The proposed problem is transformed into an optimization problem such that the L2−error remains minimal. A single hidden layer is chosen to reduce the sink of the local minimum values. The weights in the neural network are trained with a hybrid cuckoo search algorithm to refine them so that we obtain a better approximate solution for the given problem. To show the efficacy of our method, we considered six different corneal models. For validation, the solution with Adam’s method is taken as a reference solution. The results are presented in the form of figures and tables. The obtained results are compared with the fractional order Darwinian particle swarm optimization (FO-DPSO). We determined that results obtained with BHCS-ANN outperformed the ones acquired with other numerical routines. Our findings suggest that BHCS-ANN is a better methodology for solving real-world problems.
... Particle accelerators, such as the spallation neutron source (SNS) (Henderson et al., 2014) and CERN, are complex engineering systems that use electromagnetic fields to propel charged particles to very high speeds and energies to use them for fundamental research applications. The interest in machine learning for control applications in particle accelerators can be seen in these studies (Nguyen, Lee, Sass, & Shoaee, 1991;Edelen et al., 2016). Uncertainty-aware anomaly detection framework of the errant beam pulses was developed by (Blokland et al., 2021) using Siamese neural networks with ResNet blocks. ...
Early fault detection and fault prognosis are crucial to ensure efficient and safe operations of complex engineering systems such as the Spallation Neutron Source (SNS) and its power electronics (high voltage converter modulators). Following an advanced experimental facility setup that mimics SNS operating conditions, the authors successfully conducted 21 early fault detection experiments, where fault precursors are introduced in the system to a degree enough to cause degradation in the waveform signals, but not enough to reach a real fault. Nine different machine learning techniques based on ensemble trees, convolutional neural networks, support vector machines, and hierarchical voting ensembles are proposed to detect the fault precursors. Although all 9 models have shown a perfect and identical performance during the training and testing phase, the performance of most models has decreased in the next test phase once they got exposed to realworld data from the 21 experiments. The hierarchical voting ensemble, which features multiple layers of diverse models, maintains a distinguished performance in early detection of the fault precursors with 95% success rate (20/21 tests), followed by adaboost and extremely randomized trees with 52% and 48% success rates, respectively. The support vector machine models were the worst with only 24% success rate (5/21 tests). The study concluded that a successful implementation of machine learning in the SNS or particle accelerator power systems would require a major upgrade in the controller and the data acquisition system to facilitate streaming and handling big data for the machine learning models. In addition, this study shows that the best performing models were diverse and based on the ensemble concept to reduce the bias and hyperparameter sensitivity of individual models.