[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-81866-en":3,"doc-seo-81866-105":29,"detail-sidebar-cat-0-en-105":94},{"code":4,"msg":5,"data":6},0,"success",{"doc_id":7,"user_id":8,"nickname":9,"user_avatar":10,"doc_module":4,"category_id":11,"category_name":12,"doc_title":13,"doc_description":14,"doc_content":15,"file_id":16,"file_url":17,"file_type":18,"file_size":19,"view_count":4,"is_deleted":4,"is_public":20,"is_downloadable":20,"audit_status":20,"page_count":21,"language":22,"language_code":23,"site_id":24,"html_lang":23,"table_of_contents":25,"faqs":26,"seo_title":13,"seo_description":14,"update_tm":27,"read_time":28},81866,2336464648322,"Aria","https://ap-avatar.wpscdn.com/avatar/2200025388227c56fec?_k=1778556882303663488",8,"Research & Report","RTL Fault Injection of a Deployed Graph Neural Network Trigger for Belle II","As particle physics detectors expand, High Energy Physics experiments must process larger data streams under strict Level-1 latency constraints. This study introduces the first Register Transfer Level fault-injection analysis of the deployed hardware neural-network trigger GNN-ETM in the Belle II system. Single-Event Upsets are injected to cover three high-impact failure modes—deadlocks, timeouts, and packet-integrity violations. Results reveal monitoring asymmetry and show inter-stage liveness monitoring yields Mean Time To Failure estimates differing by up to 78.7%, enabling stage-level hardening priorities.","RTL Fault Injection of a Deployed Graph Neural Network Trigger for Belle II  \nGeorgios Sotiropoulos∗ , Marc Neu∗ , Tanja Harbaum∗ , Torben Ferber∗ , J¨urgen Becker∗  \n∗ Karlsruhe Institute of Technology (KIT)  \nEmail: {georgios.sotiropoulos, marc.neu, harbaum, torben.ferber, [becker](becker}@kit.edu)[}](becker}@kit.edu)[@kit.edu](becker}@kit.edu)  \narXiv :2607 .03 153v 1 [hep-ex] 3 Jul 2026  \nAbstract—As particle physics detectors grow in scale, High Energy Physics experiments must process ever-increasing data volumes. Level-1 trigger systems, implemented on FieldProgrammable Gate Arrays and increasingly using neuralnetwork algorithms, filter this data in real time. However, their proximity to the interaction point exposes them to radiation, which can corrupt outputs, stall processing pipelines, or damage hardware, with significant financial and scientific consequences. In this work, we present the first Register Transfer Level faultinjection study of a deployed Level-1 hardware neural-network trigger, GNN-ETM in the Belle II trigger system. We target three failure modes most consequential to a real-time trigger pipeline: deadlocks, timeouts, and packet-integrity violations. Through two complementary campaigns, we inject 1442840 Single-Event Upsets across 211245 signals. We find a monitoring asymmetry in the existing verification infrastructure and propose inter-stage liveness monitoring as a more accurate alternative to output-only observation, showing that Mean Time To Failure estimates from the two approaches differ by up to 78.7% . The resulting per-stage data identifies the highest-priority hardening targets.  \nI. INTRODUCTION  \nModern particle physics detectors have grown significantly and are capable of generating extremely large volumes of data. To reduce the amount of data for offline storage, High Energy Physics (HEP) experiments employ trigger systems, which are responsible for filtering, in real time, the data to be saved for further analyses. Because trigger applications must satisfy strict latency and throughput constraints, they are typically implemented on Field-Programmable Gate Arrays (FPGAs) .  \nIn recent years, Neural Network (NN) models have been increasingly integrated into trigger systems for a variety of tasks in several experiments, such as Belle II [1]–[3] and CERN Compact Muon Solenoid (CMS) detector [4], [5] . However, as these NN-based trigger systems are deployed close to the interaction point, they are exposed to radiation, caused by particle collisions in the detector [6] .  \nRadiation-induced faults in hardware can lead to corrupted output data, stalled processing pipelines or even cause permanent hardware damage [7] . Given the scale of investment, failures in the trigger logic that cause downtime can result insubstantial financial and scientific costs.  \nPrior fault-injection studies on NN accelerators have concentrated on Silent Data Corruption (SDC), while Detected Unrecoverable Errors (DUEs) (hangs, deadlocks, timeouts) remain understudied [8] . For a real-time trigger, DUEs and packet-integrity violations are the most consequential failures. Existing studies also evaluate benchmark workloads on rep-  \nFig. 1. Overview of the RTL fault injection testbench used to evaluate the reliability of GNN-ETM.  \nresentative accelerators rather than full end-to-end deployed systems.  \nA timely and realistic case study for a fault tolerance analysis is the Belle II trigger system, presented in [3], which is an FPGA-based Graph Neural Network (GNN) dataflow accelerator for calorimeter clustering and triggerbit generation. Its scale and complexity render traditional mitigation techniques such as Triple Modular Redundancy (TMR) impractical, since the associated area overhead would prevent the design from fitting on a single FPGA. Moreover, the latency overhead of using multiple FPGAs would violate the strict timing constraints.  \nIn this paper, we investigate how a complex hardware trigger can be affected","cbCaioWGsBQMGI2U","https://ap.wps.com/l/cbCaioWGsBQMGI2U","pdf",270140,1,6,"English","en",105,"# I. Introduction\n# II. Related Work","[{\"question\":\"What is the main purpose of the study?\",\"answer\":\"The study performs a Register Transfer Level fault-injection analysis of the deployed Belle II Level-1 neural-network trigger (GNN-ETM) to evaluate reliability under radiation-induced faults.\"},{\"question\":\"Which failure modes are targeted as most consequential?\",\"answer\":\"Deadlocks, timeouts, and packet-integrity violations are targeted because they most directly disrupt a real-time trigger pipeline.\"},{\"question\":\"How is the proposed monitoring approach different from existing verification?\",\"answer\":\"The work proposes inter-stage liveness monitoring as a more accurate alternative to output-only observation, addressing a monitoring asymmetry in the verification infrastructure.\"},{\"question\":\"What impact do the monitoring methods have on MTTF estimation?\",\"answer\":\"Mean Time To Failure estimates derived from the two approaches differ by up to 78.7%, and the resulting per-stage data identifies the highest-priority hardening targets.\"}]",1784176736,15,{"code":4,"msg":30,"data":31},"ok",{"site_id":24,"language":23,"slug":32,"title":13,"keywords":33,"description":14,"schema_data":34,"social_meta":89,"head_meta":91,"extra_data":93,"updated_unix":27},"rtl-fault-injection-of-a-deployed-graph-neural-network-trigger-for-belle-ii","",{"@graph":35,"@context":88},[36,53,67],{"@type":37,"itemListElement":38},"BreadcrumbList",[39,43,47,50],{"item":40,"name":41,"@type":42,"position":20},"https://docshare.wps.com","Home","ListItem",{"item":44,"name":45,"@type":42,"position":46},"https://docshare.wps.com/document/","Document",2,{"item":48,"name":12,"@type":42,"position":49},"https://docshare.wps.com/document/research-report/",3,{"item":51,"name":13,"@type":42,"position":52},"https://docshare.wps.com/document/rtl-fault-injection-of-a-deployed-graph-neural-network-trigger-for-belle-ii/81866/",4,{"url":51,"name":13,"@type":54,"author":55,"headline":13,"publisher":57,"fileFormat":60,"inLanguage":23,"description":14,"dateModified":61,"datePublished":61,"encodingFormat":60,"isAccessibleForFree":62,"interactionStatistic":63},"DigitalDocument",{"name":9,"@type":56},"Person",{"url":40,"name":58,"@type":59},"DocShare","Organization","application/pdf","2026-07-16",true,{"@type":64,"interactionType":65,"userInteractionCount":4},"InteractionCounter",{"@type":66},"ViewAction",{"@type":68,"mainEntity":69},"FAQPage",[70,76,80,84],{"name":71,"@type":72,"acceptedAnswer":73},"What is the main purpose of the study?","Question",{"text":74,"@type":75},"The study performs a Register Transfer Level fault-injection analysis of the deployed Belle II Level-1 neural-network trigger (GNN-ETM) to evaluate reliability under radiation-induced faults.","Answer",{"name":77,"@type":72,"acceptedAnswer":78},"Which failure modes are targeted as most consequential?",{"text":79,"@type":75},"Deadlocks, timeouts, and packet-integrity violations are targeted because they most directly disrupt a real-time trigger pipeline.",{"name":81,"@type":72,"acceptedAnswer":82},"How is the proposed monitoring approach different from existing verification?",{"text":83,"@type":75},"The work proposes inter-stage liveness monitoring as a more accurate alternative to output-only observation, addressing a monitoring asymmetry in the verification infrastructure.",{"name":85,"@type":72,"acceptedAnswer":86},"What impact do the monitoring methods have on MTTF estimation?",{"text":87,"@type":75},"Mean Time To Failure estimates derived from the two approaches differ by up to 78.7%, and the resulting per-stage data identifies the highest-priority hardening targets.","https://schema.org",{"og:url":51,"og:type":90,"og:title":13,"og:site_name":58,"og:description":14},"article",{"robots":92,"canonical":51},"index,follow",{"doc_id":7,"site_id":24},{"code":4,"msg":5,"data":95},[96,100,104,108,113,117,122,125,130,133,137],{"id":20,"doc_module":4,"doc_module_name":45,"category_name":97,"show_sort_weight":98,"slug":99},"Story & Novel",90,"story-novel",{"id":46,"doc_module":4,"doc_module_name":45,"category_name":101,"show_sort_weight":102,"slug":103},"Literature",80,"literature",{"id":52,"doc_module":4,"doc_module_name":45,"category_name":105,"show_sort_weight":106,"slug":107},"Exam",70,"exam",{"id":109,"doc_module":4,"doc_module_name":45,"category_name":110,"show_sort_weight":111,"slug":112},5,"Comic",60,"comic",{"id":21,"doc_module":4,"doc_module_name":45,"category_name":114,"show_sort_weight":115,"slug":116},"Technology",50,"technology",{"id":118,"doc_module":4,"doc_module_name":45,"category_name":119,"show_sort_weight":120,"slug":121},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":45,"category_name":12,"show_sort_weight":123,"slug":124},30,"research-report",{"id":126,"doc_module":4,"doc_module_name":45,"category_name":127,"show_sort_weight":128,"slug":129},9,"Religion & Spirituality",20,"religion-spirituality",{"id":128,"doc_module":4,"doc_module_name":45,"category_name":131,"show_sort_weight":128,"slug":132},"World Cup","world-cup",{"id":134,"doc_module":4,"doc_module_name":45,"category_name":135,"show_sort_weight":134,"slug":136},10,"Lifestyle","lifestyle",{"id":138,"doc_module":4,"doc_module_name":45,"category_name":139,"show_sort_weight":109,"slug":140},19,"General","general"]