Holger Schlarb

Jan Kaiser, Chenran Xu, Annika Eichler et al.

Online tuning of real-world plants is a complex optimisation problem that continues to require manual intervention by experienced human operators. Autonomous tuning is a rapidly expanding field of research, where learning-based methods, such as Reinforcement Learning-trained Optimisation (RLO) and Bayesian optimisation (BO), hold great promise for achieving outstanding plant performance and reducing tuning times. Which algorithm to choose in different scenarios, however, remains an open question. Here we present a comparative study using a routine task in a real particle accelerator as an example, showing that RLO generally outperforms BO, but is not always the best choice. Based on the study's results, we provide a clear set of criteria to guide the choice of algorithm for a given tuning task. These can ease the adoption of learning-based autonomous tuning solutions to the operation of complex real-world plants, ultimately improving the availability and pushing the limits of operability of these facilities, thereby enabling scientific and engineering advancements.

Leandro Lanzieri, Peter Kietzmann, Goerschwin Fey et al.

Ageing detection and failure prediction are essential in many Internet of Things (IoT) deployments, which operate huge quantities of embedded devices unattended in the field for years. In this paper, we present a large-scale empirical analysis of natural SRAM wear-out using 154 boards from a general-purpose testbed. Starting from SRAM initialization bias, which each node can easily collect at startup, we apply various metrics for feature extraction and experiment with common machine learning methods to predict the age of operation for this node. Our findings indicate that even though ageing impacts are subtle, our indicators can well estimate usage times with an $R^2$ score of 0.77 and a mean error of 24% using regressors, and with an F1 score above 0.6 for classifiers applying a six-months resolution.

Leandro Lanzieri, Jiri Kral, Goerschwin Fey et al.

Microcontrollers are increasingly present in embedded deployments and dependable systems, for which malfunctions due to hardware ageing can have severe impact. The lack of deployable techniques for ageing monitoring on these devices has spread the application of guard bands to prevent timing errors due to degradation. Applying this static technique can limit performance and lead to sudden failures as devices age. In this paper, we follow a software-based self-testing approach to design monitoring of hardware degradation for microcontrollers. Deployable in the field, our technique leverages timing windows of variable lengths to determine the maximum operational frequency of the devices. We empirically validate the method on real hardware and find that it consistently detects temperature-induced degradations in maximum operating frequency of up to 13.79 % across devices for 60 °C temperature increase.

Jun Zhu, Ye Chen, Frank Brinker et al.

Modeling of large-scale research facilities is extremely challenging due to complex physical processes and engineering problems. Here, we adopt a data-driven approach to model the longitudinal phase-space diagnostic beamline at the photoinector of the European XFEL with an encoder-decoder neural network model. A deep convolutional neural network (decoder) is used to build images measured on the screen from a small feature map generated by another neural network (encoder). We demonstrate that the model trained only with experimental data can make high-fidelity predictions of megapixel images for the longitudinal phase-space measurement without any prior knowledge of photoinjectors and electron beams. The prediction significantly outperforms existing methods. We also show the scalability and interpretability of the model by sharing the same decoder with more than one encoder used for different setups of the photoinjector, and propose a pragmatic way to model a facility with various diagnostics and working points. This opens the door to a new way of accurately modeling a photoinjector using neural networks and experimental data. The approach can possibly be extended to the whole accelerator and even other types of scientific facilities.