SUMMARY AND CONCLUSIONSThis paper studies the vibration signals obtained from a reactor coolant pump (RCP) of a power plant. A discrete wavelet transform (DWT) is used to decompose the vibration signals generated within a time window into approximation (cA2) and detailed coefficients (cD2 and cD1). Additionally, twelve features consisting of some statistics are generated from each of the DWT coefficients or sub-bands, resulting in a total of 36 features per vibration signal. These features are used as inputs for the principal component analysis (PCA). From the PCA results, only 4 principal components (PCs) explain most variability (about 96%) of the features. The variability was found in features X2, X13, X14, X25, and X26 which correspond to the 75th percentile of cA2, 25th percentile of cD2, 75th percentile of cD2, 25th percentile of cD1, and 75th percentile of cD1 respectively. Therefore, we can monitor these small number of 4 PCs variables instead of the original 36 PCs. These 4` PCs can provide the behavior of normal vibration signals and how to detect abnormal signals in the future.
SUMMARY & CONCLUSIONSThis paper describes the results of a survey that was conducted to determine the knowledge and skills that reliability engineering practitioners utilize in their everyday work. The survey consisted of twenty-two (22) questions and was conducted in person at two (2) events. The first event was the 13th Annual Training Summit event held by the Society of Reliability Engineers (SRE) – Huntsville Chapter, in Huntsville, Alabama on November 29-30, 2021. The second event was the 68th Annual RAMS in Tucson, Arizona on January 24-27, 2022. In all, there were fifty (50) respondents to the survey with two being determined ineligible for lack of working in a reliability engineering role. Results showed that the vast majority of reliability engineers who participated in the survey were interested in having better data science skills for themselves and were also interested in having a teammate with data science skills. Approximately 56% of practitioners that were surveyed showed preference for having a reliability engineer with data science skills on their team. By contrast, 26% preferred a data scientist with reliability engineering knowledge and 14% had no preference. Survey results also showed that the data science skills used most frequently by reliability engineers are statistical summaries and analysis, data modeling, predictive modeling, data visualization, data cleansing/preparation, and data mining. Finally, survey results showed not only an interest hiring reliability engineers with data science skills, but also a significant openness to hiring data scientists with reliability engineering skills.
SUMMARY & CONCLUSIONSThis paper provides a Reliability & Maintainability Strategic Decision-Making Framework Model for repairable systems in Rail Transportation Fleets. The Model is supported by, a total of 84 case studies, over a period of fifteen years, suggests that Safety is a function of Reliability. Those case studies on a large series of nonstructural repairable systems and components, impacting customer service in subway transportation fleets demonstrate that a greater proportion of the items examined exposed non-ageing related failure patterns over their in-service life-cycles. These results of In-service incursions, modeled by the use of probability distributions widely used in lifetime series analysis, revealed wear out patterns in agreement with studies in aircraft and ship fleets, as documented in Nowlan & Heap Reliability Report for UAL 1968 [1]. Remarkably, the distinctiveness of patterns which allow for the characterization of Age-Related from Non-Ageing related failure distributions provide a strategic decision making framework for component maintainability and asset configuration performance optimization within transportation operation system safety and cost. The project revealed relevant and significant truths and miss concepts in asset maintenance management reliability, such as: If an asset repair is not properly described, diagnosed and prescribed and executed, it could cause more negative impact in customer service, frustrating the expectation of a reliable performance. Also, if an asset that doesn’t expose typical age-related wear out failure distribution pattern is placed under a scheduled time-based maintenance repair or overhauls, it will not or marginally benefit from the program work; inevitably can re-introduce "infant-mortality" or premature failures. Results suggest that maintenance doesn’t improve asset reliability, it is meant to restore an item functions as close as new possible, however it is observed that there is always decay and problems of obsolescence. We argue that that results presented in this paper are a novel practical quantitative confirmation of the Reliability Centered Maintenance (RCM) [1] failure patterns in rail mass transit systems with Subway Failure Patterns in agreement with RCM and Failures in Aircrafts and Ships. The confirmation establishes a foundation for an innovative Reliability & Maintainability Strategic Decision-Making Framework Model for repairable Systems in Rail Transportation Fleets, which may further offer insights on the limitation on the application of the RCM Decision Diagram when there is no sufficient data to support an analysis. As Safety, Reliability & Maintainability are interconnected and cannot be separated; the Model can support safety analysis. With the exponential increase in technology complexity on the application of "Fault-Tolerant" & "Fail-Safe" designs, Safety is still a function of Reliability. Therefore, due to the fact we cannot completely dissociate the much close relationship between Reliability & Safety the Model proves to be profoundly essential to support the Safety Culture.
Summary and ConclusionsThis paper presents a methodology to show compliance to safety objectives using a new data curation system called RACK. First, we use a bowtie diagram as a lightweight safety analysis to demonstrate the connection between safety assessment and software development process. Once an assurance level is properly assigned, we use a Python script to highlight which DO-178C objectives we are obligated to show compliance to. Next, we present the use of RACK—Rapid Assurance Curation Kit—to find evidence in each of the software development processes. Finally, we present a novel way to concisely visualize traceability and status of compliance to DO-178C objectives via a sunburst. Our finding is that provenance language is embedded in DO-178C and their referenced activities. This makes RACK suitable for demonstrating compliance since its underlying data model is built on provenance classes and relationships. In summary, this paper presents a methodology that automatically generates a visualization to show compliance to DO-178C objectives using RACK to query for evidence that an activity was completed.
SUMMARYWhile the definition of a Digital Twin (DT) is provided within recent literature, each DT should be designed to meet the specific requirements of the user. As RAMS is focused on the identification, understanding and mitigation of technical risk in a system, a RAMS engineer requires a DT that can autonomously establish the potential dependencies and impacts of functional and physical failures on a system, and auto-generate various analyses to identify the appropriate mitigation approach.The concept of a Digital Risk Twin (DRT) described in this paper should uses an integrated and inter-related set of information about the system, autonomously reflecting changes across analyses that utilize causal simulation to understand potential risks and map their dependencies. This definition has been examined thoroughly from the original literature for Digital Twins, through an examination of core system safety/risk assessment practices demonstrated in RAMS and finally arrives at the main point of the definition and key aspects of a Digital Risk Twin (DRT).As the DRT digitizes the RAMS process across each stage of the Product Lifecycle, it is important that it offers common DT features such as integration, visualization, and simulation. The DRT will also implicitly digitize the engineering domain knowledge utilized in the design process (‘Digital Domain Knowledge’), providing a persistent context for analysis and design decisions, and so requires a framework of automated data management, maintaining traceability between activities and decisions made in relation to identified risks.
SUMMARY & CONCLUSIONSIn this work, a physics-based reliability model (PFKS model) of HfO x based resistive random-access memory (RRAM) is developed using Kinetic Monte Carlo (KMC) simulation with MATLAB to demonstrate the correlation between microscopic kinetics, behavioral degradation and variation of RRAM. The resistive switching process of RRAM is simulated through a dynamic resistor network. The physics of oxygen vacancy generation, oxygen ion migration, hopping and recombination is modeled through multiple kinetic processes during switching, including forming, SET, and RESET. The electric field, current temperature distribution, resistance and microstructural development of the dielectric are analyzed simultaneously and updated in time steps. The proposed PFKS model looks into the physical causes of endurance and retention degradations as a function of temperature, number of cycles, and time. The PFKS model is capable of simulating operation cycles, with flexible setups in material properties (adaptable to new materials combinations), structures (adaptable to various setups in device layers), and applied conditions (regarding temperature, voltage profiles, and compliance current).This paper describes the motivation for studying RRAM, and the need for looking into RRAM reliability attributes and characteristic for trade-offs. The workflow and simulation set-ups of this PFKS model are demonstrated schematically and mathematically. The simulation results for retention and endurance are discussed. The degradation rate of RRAM can be further modified by constructing a Bayesian network of reliability attributes and assigning conditional dependencies. In summary, the proposed PFKS model provides insights on trade-offs among materials characteristics, structure differences, and operating conditions considering RRAM operation and reliability optimization.
In this paper, we propose a Machin Learning (ML)-based framework for maintenance decision making for multi-unit system. More specifically, we propose Reinforcement Learning (RL) approach for dynamic maintenance model for multi- component parallel system subject to stochastic degradation and random failures. Deterioration of each unit occurs independently according to a three-state homogenous Markov process such that each unit has three states, namely, healthy, unhealthy, and failure state. The interaction among system states are modeled based on Birth/Birth-Death process. The overall system state is defined based on different combination of individual component state. The optimal maintenance policy for the system is obtained by modeling the problem as Markov Decision Process (MDP) and Q-learning algorithm with focus on cost minimization is applied as a solution methodology. In comparison to tradition MDP approaches, proposed RL solution is more effective and practical in terms of time and cost savings. Specifically, when the state-space of the problem is large, traditional MDP is note capable to converge to the optimal policy in a timely fashion. Therefore, there is a is the decisive need for development of RL-based solution for maintenance decision making. A numerical example is provided which demonstrates how the RL can be used to find the optimal maintenance policy for the system under study.
SUMMARY & CONCLUSIONSThis paper presents a Design for Availability approach for repairable health care Systems and Solutions; based on a framework which merges modeling approaches with managing and balancing of Reliability and Maintainability characteristics. The approach is used to prioritize the required Design for Reliability and Maintainability activities during the development phases of health care Systems and Solutions. It is supported by an ‘easy’ and ‘intuitive’ web-based toolset.The presented approach is applicable to product development and continuous improvement of fielded systems. It also serves as an enabler of digital twins for further Availability optimization.The Design for Availability approach starts with modeling of the System and Solution design in the Availability Design Tool, considering existing and new functionalities as well as different maintenance types. This is done in a hierarchal way to manage complexity. Historical data analysis as well as change point analysis are used to predict the Reliability, Availability, Maintainability and Life Cycle Cost of the Systems and Solutions under development.Field data on both Reliability as well as Maintainability aspects are obtained via a data pipeline, thereby connecting the web-based Availability Design Tool with the data lake. For non-part related Maintenance activities, work order text analytics with Deep Learning algorithms is used to classify the activities and their impact. This is fed to the Availability model to improve the overall predictive capability of the model.With the Availability model and FMEA as inputs, a qualitative and quantitative analysis is executed to determine the Maintainability and Reliability of critical elements and parts in the proposed design. Easy execution of the FMEA is supported by a web based FMEA tool, allowing clear technical risk identification and prioritization.Based on the criticality analysis, work packages are defined. The work packages contain the Design for Reliability and Maintainability activities required, to ensure that the Systems and Solutions Availability and Life Cycle Cost objectives are met.Upon execution of the Design for Reliability and Maintainability activities, the initial model will require updates because of improved knowledge and understanding of probabilities of failure modes. This is also necessary due to the implemented design changes which mitigate the identified Reliability and Maintainability criticalities; either by preventing failures from occurring, implementation of fault tolerance or fault removal strategies.The approach allows for creating a hierarchical Availability model which includes Hardware as well as Software elements and feeding it with (near real-time) field data. This allows and prepares the model for use as digital twin for further optimization of the Availability of health care Systems and Solutions. The approach can be applied on an individual system as well as on multiple systems; enabled by an easy and free-to-use web-based toolset.
Summary & ConclusionsResilience, a system property merging the consideration of stochastic and malicious events focusing on mission success, motivates researchers and practitioners to develop methodologies to support holistic assessments. While established risk assessment methods exist for early and advanced analysis of complex systems, the dynamic nature of security is much more challenging for resilience analysis.The scientific contribution of this paper is a methodology called Trust Loss Effects Analysis (TLEA) for the systematic assessment of the risks to the mission emerging from compromised trust of humans who are part of or are interacting with the system. To make this work more understandable and applicable, the TLEA method follows the steps of Failure Mode, Effects & Criticality Analysis (FMECA) with a difference in the steps related to the identification of security events. There, the TLEA method uses steps from the Spoofing, Tampering, Repudiation, Information disclosure, Denial of Service (DoS), Elevation of privilege (STRIDE) methodology.The TLEA is introduced using a generic example and is then demonstrated using a more realistic use case of a drone-based system on a reconnaissance mission. After the application of the TLEA method, it is possible to identify different risks related to the loss of trust and evaluate their impact on mission success.
SUMMARY & CONCLUSIONSThis work introduces a systematic approach to model the degradation of the capacity of lithium-ion (li-ion) batteries in Electronic Heating Systems (EHS) products and relates the model to field capacity degradation based on the users’ usage pattern. The current literatures on battery capacity estimation focus on the mathematical evaluation based on electrochemical, semi-empirical, and data-driven models. However, these approaches lack the connection to predicting battery degradation based on actual usage pattern in the field.By utilizing the Accelerated Life Test (ALT) data and battery knowledge, the proposed methodology facilitates the integration of non-linear regression-based Transfer Function (TF) with Monte Carlo simulation for the prediction of the battery capacity degradation. This enables the battery system designers to make data-driven decisions.The method is applied to small hand-held li-ion battery-driven EHS products. Parameters affecting the battery capacity degradation of such devices include continuous usage with intermittent resting before charging, variable resting time, and different usage durations. The TF helps in relating these critical parameters to the battery capacity degradation. The outcome of the methodology is then compared with the actual field performance where a good approximation is observed.Historically, Philip Morris Products (PMP) SA used usage cycle and linear-regression model to predict battery capacity degradation based on the worst-case users. By implementing this non-linear modeling approach integrated with Monte Carlo methodology, it is possible to fit the field usage pattern (month-to-month) to precisely predict the device usage limitation constrained by battery degradation. This was combined with an extrapolation to predict a possible warranty extension.
Summary & ConclusionsCorrosion defect in oil and gas pipelines is the major risk factor that threatens the structural integrity of buried pipelines. Pipeline corrosion management which typically requires high-resolution inline inspections (ILI) to characterize the corrosion growth process is an important task to assess the corrosion defects. However, the corrosion process is inherently stochastic, temporal-spatial dependent, and driven by various hidden physics-related variables. In the literature, the growth of corrosion defects in pipelines is usually modeled by various stochastic processes imposing many unjustified assumptions about the mean growth path and probability distribution of the corrosion rate. In this paper, we proposed a physics-informed latent variable corrosion growth model that integrates the known physics from complex processes into modeling the relations between observable and latent variables and the actual stochastic process that generate the corrosion time series data. The proposed method consists of 3 main steps. Firstly, the latent and observed variable relations and the underlying stochastic processes governing the corrosion process were modelled by physics-informed regression models. The latent groupings that give rise to the observed time series were identified by using the agglomerative hierarchical clustering method. Finally, the prediction algorithm based on the learned physic-informed regression model was proposed to forecast the process progression. We validated the model by a case study on a simulated corrosion process in oil and gas pipelines, in which both latent variables and ILI data were sampled from predefined distributions. The results indicated that the model could predict the growth of corrosion defects and capture the variance of the stochastic processes demonstrated by the low mean absolute percentage errors (MAPE) of 3.0082%, 3.9532%, and 3.6831%, which corresponds to the three corrosion growth processes causes by three types of soil. The proposed model can be used to facilitate the development of reliability models and corrosion management.
Summary & ConclusionsThe use of composite pipelines has grown in recent years as an alternative to steel pipelines in the offshore oil and gas industry, due their excellent corrosion resistance and high specific strength. Among these developments, thermoplastic composite pipes have seen a recent rise in research and use due to their greater flexibility and impact resistance than traditional thermoset-matrix composites. To understand the performance of these materials in the field, studies have derived experimental, analytical, and theoretical estimates of the mechanical and thermal properties of TCP in dry conditions, but studies in wet or acidic environments are currently restricted to the empirical domain. As a part of ongoing work to predict behavior in these conditions, this study describes a method to predict the degradation of mechanical properties from one form of environmental degradation: diffusion of the transport fluid through the pipe wall. This process forms a gel layer at the interface between the pipe and fluid. The thickness of the gel layer and degree of stiffness reduction within the gel are taken as metrics for pipeline degradation. A test case of a polypropylene/E-glass fiber (PP/G) laminate pipe is presented to show the effects of this degradation on the pressure capacity of the pipe. It was found that as the gel layer thickness approaches that of the innermost ply, the internal pressure capacity is significantly reduced.
SUMMARY & CONCLUSIONSOne essential task in practice is to quantify and improve the reliability of an infrastructure network in terms of the connectivity of network components (i.e., all-terminal reliability). However, as the number of edges and nodes in the network increases, computing the all-terminal network reliability using exact algorithms becomes prohibitive. This is extremely burdensome in network designs requiring repeated computations. In this paper, we propose a novel machine learning-based framework for evaluating and improving all-terminal network reliability using Deep Neural Networks (DNNs) and Deep Reinforcement Learning (DRL). With the help of DNNs and Stochastic Variational Inference (SVI), we can effectively compute the all-terminal reliability for different network configurations in DRL. Furthermore, the Bayesian nature of the proposed SVI+DNN model allows for quantifying the estimation uncertainty while enforcing regularization and reducing overfitting. Our numerical experiment and case study show that the proposed framework provides an effective tool for infrastructure network reliability improvement.
SUMMARY & CONCLUSIONSThis paper presents a methodology for applying topic modelling and deep learning to unstructured maintenance data, for improving availability and cost savings. Learning from textual technical data requires special skills and time. As a result, the valuable insights in such data are often trapped and not fully harnessed when necessary.The approach demonstrated in this work takes texts from service records as input to a Bidirectional Encoder Representations from Transformers (BERT) model, and yields service action topics, according to frequency of occurrence. The yielded service action topics were regarded as failures and linked to the timestamps when they occurred. Failures were selected and analyzed using a 3-parameter Weibull model to determine the reliability profile. The Weibull model parameters from the model were used to determine the optimal repair time, considering the theoretical costs of preventative and corrective maintenance. Additionally, deep learning was applied to the service diagnostic texts to classify them into preventative maintenance and corrective maintenance.The benefit of the approach lies on its applicability to large textual datasets; to identify recurring activities hidden in maintenance records, prioritize cost savings and digitally transform system availability.
Summary & ConclusionsTo aid and improve the reliability of product designs, repeated safety tests are required to find out the safety performance of the product with respect to design variables. A large number of design variables involved in the performance evaluations often leads to enormous testing costs. A method that can effectively utilize partially available information from multiple sources of varying dimensions and fidelity is a pressing need for reliability-based product design. Moreover, in the product design and safety estimation process, it is beneficial to take into account the manufacturing policies and physical principles. Therefore, it is desirable to have a framework that allows the incorporation of physical principles and other prior information to regularize the behavior of the predictive model. This paper presents a new physics-constrained machine learning method for reliability-based product design and safety estimation considering partially available limited reliability information.
SUMMARY & CONCLUSIONSLevel 4 (L4) Automated Driving Systems (ADS) development and testing efforts target not only personal use, but also its deployment within a Mobility as a Service (MaaS) context. MaaS operations introduce additional safety aspects related to passenger behavior, fleet operators, communication between fleet operators, ADS developers, and vehicle manufacturers. However, the operational safety of L4 ADS vehicles operating within a MaaS context has not been fully addressed.The Society of Automotive Engineers (SAE) J3016 standard provides a solid foundation for the concepts related to ADS operations, including safety-related ones such as how the ADS vehicle should behave in case of a system failure or when exiting the domain in which it is designed to operate – DDT Fallback. This paper discusses J3016 essential concepts and the challenges posed by the use of ADS vehicles, within a MaaS operation, to some of these concepts. Through Event Sequence Diagrams, new goals for the ADS DDT fallback are proposed, and the roles of remote monitoring operators for safety are discussed.
SUMMARY & CONCLUSIONSDegradation methods have proven to be effective to estimate lifetime and reliability, especially for highly reliable products. To model degradation behavior, most of the earlier work has assumed degradation considering single failure processes and constant operating conditions. Nevertheless, in reality, the product may be subjected to multiple non-linear failure processes that influence the propagation of each other causing acceleration of the failure processes. Moreover, a product may be subjected to important variations in operational conditions due to a variety of environmental situations and user profiles. In addition, most of the previous work considers a single degradation characteristic and constant operating conditions. As an alternative, to deal with complex products that exhibit multiple degradation mechanisms, data-driven approaches have received attention recently. For instance, artificial neural networks (ANNs) have emerged as a promissory tool to estimate reliability. Traditional, ANNs, have evolved to deep learning (DL) approaches such as Deep Neural Networks (DNNs), convolutional neural networks (CNNs), and recurrent neural network (RNNs). Such DL approaches have been applied for reliability estimation. However, little evidence is available of Long Short-Term Memory Networks (LSTMNs) for reliability estimation. Therefore, we propose a preliminary development of product reliability estimation based on LSTMNs, a state-of-the-art deep learning technique. LSTMNs are appropriate for time-to-failure prediction as they have the capability of making regression based on sequence input data. Consequently, multidimensional degradation data can be used as input to predict the time-to-failure for later reliability estimation. To use LSTMNs for product reliability estimation, an applicable architecture and general framework needs to be developed. Different architectures are investigated. The best architecture configuration is selected based on the root mean squared error (RMSE). The proposed methodology is demonstrated with the Turbofan Engine Degradation Simulation Data Set. Although the accuracy is compromised due to the approximation nature of the proposed approach, one of its major contributions is the possibility of estimating the reliability given sensor data under any operating conditions and based on multivariate degradation data.
SUMMARYThis paper focuses on predicting lithium battery capacity and state of health based on pre-recorded datasets. Data-driven techniques, including machine learning approaches have become prevalent. However, data is often inputted into these methods without much thought. Therefore, data pre-processing before utilizing machine learning techniques can optimize the performance and efficiency of these models. Since the battery degradation process is highly nonlinear and influenced by multiple factors, including both manufacturing aspects and operating conditions, which complicate the battery capacity prognostics, we need some tool that can keep track on useful data and produce reliable predictions in the end [1]. Specifically, the main methodology of this study includes defining indirect health indicators as inputs into two machine learning techniques: Gaussian Process Regression (GPR) and Feed Forward Neural Network. To demonstrate this method, the NASA battery dataset [2] is utilized; in particular, the indirect health indicators are extracted from the battery’s temperature and charging state (the voltage and current change process), then a machine learning model is utilized to optimize prediction of capacity degradation.The first method for predicting the State of Health (SOH) of a battery is Gaussian Process Regression, which modifies the isotropic squared exponential kernel with an automatic relevance determination structure. The entire process extracts the highly relevant input features for capacity predictions [1]. Utilizing this method enables quick extraction of relevant data and provides fast predictions of capacity.The second strategy is Feedforward Neural Networking (FFNN). Overall, FFNN is a widely used machine learning method with outstanding performance in nonlinear modeling. It first intake one input as the starting layer, and each layer except the output layer is fully connected to the next layer, so the final prediction of FFNN is tightly connected with the previous data [3].However, the inputs to these advanced models need to be considered, as irrelevant inputs can lead to inaccurate predictions (see figure 1). Therefore, indirect health indicator (IHI) is applied as a form of feature extraction as an input to the model. Indirect health indicator is what affects the capacity of a battery during the charging or discharging state. During those processes, the performance of lithium-ion batteries will deteriorate with capacity decreasing and impedance increasing, which will cause equipment and system failures or even catastrophic loss [4]. Therefore, it is necessary to consider the IHI before predicting the model.For this study, indirect health indicators are determined with data pre-processing and are subsequently inputted into the GPR and FFNN. The results show success for capacity estimation for lithium ion batteries.
SUMMARY & CONCLUSIONSThe working principle of EVs is mainly based on the rechargeable battery packs which provide the power to the vehicle. Keeping these battery packs within optimal temperature is very much essential to achieve the best performance, gain autonomy (higher mileage) and preserve the battery components from deterioration. Therefore, the battery cooling system plays a fundamental role in optimizing the thermal management of battery packs. It also has implications for vehicle safety, because a failure of battery coolers may result in hazardous conditions. Thus, the reliability of battery coolers becomes an important and unavoidable task for the industries leading towards mass production manufacturing of EVs.The reliability analysis of automotive parts is usually performed at component level. In the case of battery coolers, the reliability predictions need to take into account the system-wise structure of the part, which is made by different coolers with different design. Indeed, this poses many challenges, since all possible wear-out failure modes (FMs) may have conditional dependability on each other. For example, corrosion phenomena are dependent on the position of the coolers (inside or outside the battery pack), and in the case of vibration loading how each cooler, pipes and connections are integrated within the battery pack.In the present work we have conducted an analysis considering the predominant FMs of battery coolers as dependent and modelled them through a Bayesian network (BN) approach to calculate the probability of field failure (part per million, PPM) in case of extension of the warranty period. The results obtained through the BN analysis are then compared with a simplified PPM calculator tool, based on an empirical approach, used within the company for preliminary risk assessment during the early stages of product development.We also illustrate how BN method can be applied to the risk assessment by considering the proposed design, its integration within the battery pack, and the severity of the validation tests requested by the customers. The final results show that the PPM calculator (the present method) overestimates the PPM whereas the BN provides more realistic PPM figures since it considers the dependability of FMs. The study also shows that the empirical method is simple but brings more conservative results. In conclusion, a risk assessment based on the BN has proved to be a modern and robust method to predict the reliability of mechanical components undergoing wear-out field failures.
Summary & ConclusionsSafety cases are required by several functional safety standards, specifications, and guidelines. Cybersecurity cases have recently been required by ISO/SAE 21424:2021 for automotive and EN TS 50701:2021 for the railway domain. In this paper we discuss cybersecurity cases and suggest using the topics and structure for a cybersecurity case as described in Annex G of EN TS 50701. BSI PAS 1881:2022 requires: "Trialing organizations shall develop and publish a publicly available and accessible version of the safety case". We have already developed a "safety case for the public" [1] to ensure that (1) the public is aware that safety evidence exists, (2) they are aware of relevant safety aspects when they are passengers, and (3) the vehicle’s limitations are described transparently.Trust is a dynamic process that involves initiating and building trust, responding to violations of trust (failures), and trying to rebuild (repair) trust. The building blocks of trust are not limited to the vehicle itself but also include the embedded AI (Artificial Intelligence) and its overt function. Trust is a holistic perception of the complete service, technology, and organizations responsible for developing, implementing, and certifying an autonomous vehicle.An autonomous vehicle will need acceptance from the certification bodies and the authorities, but we also need to gain the public’s trust. Our research found that several aspects are missing in the safety and cybersecurity cases to ensure public trust.To make self-driving buses a success, they need to be considered trustworthy. Thus, we need a "Trust case" that includes evidence related to distinct trust aspects. Our literature studies, focus groups [4], and surveys found that trust and safety are not correlated. We have developed a "Trust case" to cover the factors not included in the safety and cybersecurity cases. The resulting "Trust case" approach is currently in the form of specific information topics presented in a layman form and a safety case for the public [6], and specific trust topics in [7]. Further research is necessary, related to topics such as deep learning, security, and incorrect reporting to the driver due to e.g., false positive results.
