https://www.mdu.se/

Deploying deep learning models on resource-constrained hardware like Field-Programmable Gate Arrays (FPGAs) remains challenging despite advancements in quantization techniques, which often fail to map optimally to target hardware. This study proposes an end-to-end workflow for fixed-point quantization targeting FPGA accelerators, integrating hardware emulation within quantization-aware training to bridge software-hardware co-design. This ensures quantized models are optimized for real-world deployment. Evaluations on CIFAR-10 and ImageNet datasets using ResNet and VGG models show competitive performance, with up to 2% improvement in accuracy over state-of-the-art methods. This work provides insights to resolve the discrepancies that often occur between software-based quantization and hardware deployment. Our methodology effectively bridges quantization and deployment, providing a practical solution for edge device applications.

Construction equipment has important roles in industries such as construction and mining. Any downtime because of failures increase cost. Traditional diagnostic systems detect failures only after they occur, making it difficult to take precautions and prolonging repair times. This paper is the first to address the analysis of machine learning-powered Prognostic and Health Management (PHM) systems specifically for predicting failures in diesel engine air intake systems, focusing on two common issues: air leakage and Exhaust Gas Recirculation (EGR) blockage. This study compares various machine learning and deep learning models for anomaly detection and fault classification using real-world sensor data from controlled engine tests. The results demonstrate that ensemble and neural network-based machine learning methods, such as Random Forest, XGBoost, and LSTM, achieve highly successful predictions for anomaly detection and fault classification.

Fault tree analysis is a system malfunction hazard evaluation quantitative and qualitative procedure. The method is well-known and widely used, especially in the safety systems domain, where it is a mandatory integral part of the so-called "Hazard Evaluation"documentation. This paper proposes an alternative or complementary deductive fault analysis method: it uses system topology to build a hypergraph representation of the system to identify component criticality and support loss of functionality probability evaluation. Once automated, the proposed method seems promising when the system engineers explore the different architectures. They may have indication about architecture's reliability without continuous feedback from the system safety team. The system safety team must check the solution once the engineers select the final architecture. They can also use the proposed method to validate the correctness of the fault tree analysis.

Machine Learning (ML) systems require representative and diverse datasets to accurately learn the objective task. Insupervised learning data needs to be accurately annotated, whichis an expensive and error-prone process. We present a methodfor generating synthetic data tailored to the use-case achievingexcellent performance in a real-world usecase. We provide amethod for producing automatically annotated synthetic visualdata of multirotor unmanned aerial vehicles (UAV) and otherairborne objects in a simulated environment with a high degreeof scene diversity, from collection of 3D models to generation ofannotated synthetic datasets (synthsets). In our data generationframework SynRender we introduce a novel method of usingNeural Radiance Field (NeRF) methods to capture photo-realistichigh-fidelity 3D-models of multirotor UAVs in order to automatedata generation for an object detection task in diverse environments. By producing data tailored to the real-world setting, ourNeRF-derived results show an advantage over generic 3D assetcollection-based methods where the domain gap between thesimulated and real-world is unacceptably large. In the spirit ofkeeping research open and accessible to the research communitywe release our dataset VISER DroneDiversity used in this project,where visual images, annotated boxes, instance segmentation anddepth maps are all generated for each image sample.

Industrial controllers constitute the core of numerous automation solutions. Continuous control system operation is crucial in certain sectors, where hardware duplication serves as a strategy to mitigate the risk of unexpected operational halts due to hardware failures. Standby controller redundancy is a commonly adopted strategy for process automation. This approach involves an active primary controller managing the process while a passive backup is on standby, ready to resume control should the primary fail. Typically, redundant controllers are paired with redundant networks and devices to eliminate any single points of failure. The process automation domain is on the brink of a paradigm shift towards greater interconnectivity and interoperability. OPC UA is emerging as the standard that will facilitate this shift, with OPC UA PubSub as the communication standard for cyclic real-time data exchange. Our work investigates standby redundancy using OPC UA PubSub, analyzing a system with redundant controllers and devices in publisher-subscriber roles. The analysis reveals that failovers are not subscriber-transparent without synchronized publisher states. We discuss solutions and experimentally validate an internal stack state synchronization alternative. 

In the era of multiple industry trends and new technologies, avionics systems can benefit from several innovations. The complexity of modern electronics is increasing quickly and is being introduced as never before in new applications. At the algorithm level, the use of deep neural networks helps to solve problems that were never believed to be doable before. At the architecture level, hardware artificial intelligence accelerators, embedded graphical processing units, embedded sensors, etc., make it possible to create very powerful new functions. The list of new technologies is long. Besides technical challenges, system integrity and availability must be assured when integrating these new technologies into avionics functions. In this paper, we present emerging technologies and why a system-level approach is necessary when implementing these technologies. We also introduce supporting means for design assurance and fault-tolerance techniques. We illustrate the importance of a system-level approach through an example. Our example shows that when developing functions with new technologies and fault-tolerant architectures, the system safety assessment process is crucial for properly implementing a fail-safe design. It is also challenging due to potentially new failure modes.

Control systems are often an integral part of automation solutions where high reliability is crucial due to the high cost of downtime. The risk of unplanned downtime is typically reduced with redundant solutions. Additionally, safety-critical automation functions require high-integrity controllers. Today, the prevalent redundancy solution is a standby scheme, where one active primary controller drives the process while a standby backup controller is ready to take over in case of primary failure. This redundant controller pair can consist of high - integrity controllers. The automation industry is trending towards Ethernet as the sole communication medium. Our work presents an initial study of a high-integrity realization of a redundancy failure detection mechanism that guarantees only one primary controller, even in the case of network partitioning between the redundant controller pair. The failure detection is a lease-based function that leases the primary role from a central lease broker. This work discusses a high-integrity realization of the primary redundancy role leasing. We deduce and present the high-integrity-related requirements and a high-level design as an initial step towards a high-integrity realization of the redundancy role leasing. 

Selecting the correct file system is critical for space applications where risks are present. This study systematically maps and tests Ext4 versus ZFS for onboard data processing on the iX10-100 and iX5-100 payload processors. The test sets are presented along with results on several performance metrics. The conclusion is that both ZFS and Ext4 are useful, but based on certain considerations of onboard data processing, Ext4 is better than the other.

In Machine Learning systems, several factors impact the performance of a trained model. The most important ones include model architecture, the amount of training time, the dataset size and diversity. We present a method for analyzing datasets from a use-case scenario perspective, detecting and quantifying out-of-distribution (OOD) data on dataset level. Our main contribution is the novel use of similarity metrics for the evaluation of the robustness of a model by introducing relative Fréchet Inception Distance (FID) and relative Kernel Inception Distance (KID) measures. These relative measures are relative to a baseline in-distribution dataset and are used to estimate how the model will perform on OOD data (i.e. estimate the model accuracy drop). We find a correlation between our proposed relative FID/relative KID measure and the drop in Average Precision (AP) accuracy on unseen data.

In this paper, we present a novel method for detecting UAVs using diverse parallel neural networks with re-inference. The parallel networks are of type Convolutional Neural Networks (CNNs). We first set up a lowthreshold (2 respectively 20%) for each of the individual networks to detect a flying object. If all networks detecta flying object in the same area of a video frame with some overlap, we zoom into that area and redo the objectdetection and classification (re-inference step). To ensure correctness and reliability of the results from severalparallel CNNs, we introduce total confidence Tc as a measurement. We also introduce the intersection overunion for multiple parallel networks, IoUAll , and use that as threshold for calculating a reliable Tc . The resultsshow great improvements regarding accurate detection of flying drones, reduced mispredictions of otherobjects as drones, and fast response time when drones disappear from the scene.