Developing large-scale industrial systems requires high-quality requirements to avoid costly rework and project delays. However, linguistic ambiguities in natural language (NL) requirements have been a long-standing challenge, often introducing misinterpretations and inconsistencies that propagate throughout the development lifecycle. Such ambiguous NL requirements necessitate early detection and well-reasoned explanations to clarify and prevent further misunderstandings among stakeholders. While solutions have been developed to detect ambiguities in NL requirements, the advent of generative large language models (LLMs) offers new avenues for explanation-augmented requirements ambiguity detection. This paper empirically investigates LLMs for ambiguity detection and explanation in real-world industrial requirements by adopting an in-context learning paradigm. Our results from three industrial datasets show that LLMs achieve a 20.2% average performance increase in classifying ambiguous requirements when prompted with ten relevant in-context demonstrations (10 -shot), compared to no demonstrations (0 -shot). Additionally, we conducted human evaluations of the LLM-generated outputs with eight industry experts along four dimensions-naturalness, adequacy, usefulness and relevance-to gain practical insights. The results show an average rating of 3.84 out of 5 across evaluation criteria, indicating that the approach is effective in providing supporting explanations for requirement ambiguities.

Quarry sites present unique operational challenges where the performance of heavy machinery is critical for maintaining efficiency and safety. In such environments, accurate torque prediction is essential for effective engine management and optimal task execution. This work addresses the torque prediction challenge for a wheel loader operating in quarry conditions by proposing a structured three-phase approach to feature selection that reduces model complexity while preserving predictive accuracy. In the first phase, features are selected based on domain expertise to capture the physical and operational realities of quarry machinery. A comprehensive set of features is then employed to establish a robust performance baseline. In the final phase, a data-driven analysis using SHapley Additive Explanations (SHAP) identifies the top five features that most significantly impact torque prediction. Model efficacy was validated via cross-validation, with R-squared and mean-squared error serving as the key performance indicators. Comparative analysis reveals that while SHAP-ranked features yield statistically optimal results, the expert-selected features are more aligned with the practical requirements of quarry operations. These findings support the design of efficient, interpretable digital twins for real-time decisions in challenging environments.

This paper presents an extended version of Deeper, a search-based simulation-integrated test solution that generates failure-revealing test scenarios for testing a deep neural network-based lane-keeping system. In the newly proposed version, we utilize a new set of bio-inspired search algorithms, genetic algorithm (GA), ( μ + λ ) and ( μ , λ ) evolution strategies (ES), and particle swarm optimization (PSO), that leverage a quality population seed and domain-specific crossover and mutation operations tailored for the presentation model used for modeling the test scenarios. In order to demonstrate the capabilities of the new test generators within Deeper, we carry out an empirical evaluation and comparison with regard to the results of five participating tools in the cyber-physical systems testing competition at SBST 2021. Our evaluation shows the newly proposed test generators in Deeper not only represent a considerable improvement on the previous version but also prove to be effective and efficient in provoking a considerable number of diverse failure-revealing test scenarios for testing an ML-driven lane-keeping system. They can trigger several failures while promoting test scenario diversity, under a limited test time budget, high target failure severity, and strict speed limit constraints.

This paper presents an interpretable machine-learning model for anomaly detection in door locks using torque data. The model aims to replace the human tactile sense in the quality control process, reducing repetitive tasks and improving reliability. The model achieved an accuracy of 96%, however, to gain social acceptance and operators' trust, interpretability of the model is crucial. The purpose of this study was to evaluate an approach that can improve interpretability of anomalous classifications obtained from an anomaly detection model. We evaluate four instance-based counterfactual explanators, three of which, employ optimization techniques and one uses, a less complex, weighted nearest neighbor approach, which serve as our baseline. The former approaches, leverage a latent representation of the data, using a weighted principal component analysis, improving plausibility of the counter factual explanations and reduces computational cost. The explanations are presented together with the 5-50-95 th percentile range of the training data, acting as a frame of reference to improve interpretability. All approaches successfully presented valid and plausible counterfactual explanations. However, instance-based approaches employing optimization techniques yielded explanations with greater similarity to the observations and was therefore concluded to be preferable despite the higher execution times (4-16s) compared to the baseline approach (0.1s). The findings of this study hold significant value for the lock industry and can potentially be extended to other industrial settings using timeseries data, serving as a valuable point of departure for further research.

Construction sites are moving towards using autonomous machines, such as autonomous haulers, to improve productivity and safety. However, enabling efficient and safe navigation of autonomous haulers at an open-pit mining site necessitates a dynamic map of the environment. In our previous works, we introduced a dynamic multi-layered map designed for this purpose. Subsequently, we proposed how to adopt the digital twin standard for manufacturing to implement this map. Yet, the proposed dynamic multi-layered map needs to be validated in real-world scenarios, which are not evident for such off-road domains. This paper presents an analysis of the state-of-practice scenarios used in validating current static maps for a fleet of autonomous haulers performing assigned missions in real-world open-pit mining applications. Drawing from insights from this case study and industrial expertise, this paper suggests validation scenarios for the multi-layer dynamic map. Moreover, the paper discusses simulation tools that could be utilized to assess the feasibility of dynamic maps in such off-road domains at construction sites.

Boosting the rail freight modal share is an ambitious target in Europe and North America. Yards, where freight trains are arranged, can be crucial in realizing this target by reliable dispatching to the network. This paper predicts freight train departures by developing a simulation-assisted machine learning model with two concepts: general (adding all predictors at once) and step-wise (adding predictors as they become available in sub-yard operations) for hump yards with the conventional layout to provide a generalized model for European and North American contexts. The developed model is a decision tree algorithm, validated via 10-fold cross-validation. The model's performance on three data sets-a real-world European yard, a baseline simulation, and an ultimate randomness simulation for a comparable North American yard-shows a respective R2$R<^>2$ of 0.90, 0.87, and 0.70. Step-wise inclusion of the predictors results differently for the real-world and simulation data. The global feature importance highlights maximum planned length, departure weekday, the number of arriving trains, and minimum arrival deviation as key predictors for the real-world data. For the simulation data, the most significant predictors are departure yard predictors, the number of arriving trains, and the maximum hump duration. Additionally, utilization rates-except for the receiving yard-enhance the predictions. We aim to predict freight train delay departures from the yard by implementing a simulation-assisted machine learning model via two general and step-wise concepts for including the predictors. In the general concept, we use all the predictors from yard operations at once. In the step-wise concept, the predictors are added to the model based on the stages of the operation to understand how each predictor impacts the departure delay. Our machine learning model is trained by real-world and simulation data. image

Quarry sites are complex systems that involve several heavy machines, equipment, people, and management systems working together in an unstructured off-road environment. Gaining accurate insights about these sites requires integrating models at various levels to enable a holistic view systems and processes involved and facilitate effective planning, coordination, and decision-making. In this paper, a multi-level modelling framework is proposed to provide an overall structure for the modelling of quarry sites. The motivation for this framework is drawn from insights gained through a large manufacturing company in the heavy-duty vehicle industry, providing a practical perspective on the modeling approach. The framework integrates models of different operations on site enabling effective simulation and optimization and leading to better understanding of the workflow on site and pointing out any possible bottlenecks. The feasibility of the proposed framework was validated through workshops that included a panel of experts in different areas of the field of off-road machinery production company.

[Context and Motivation] Requirements in tender documents are often mixed with other supporting information. Identifying requirements in large tender documents could aid the bidding process and help estimate the risk associated with the project. [Question/problem] Manual identification of requirements in large documents is a resource-intensive activity that is prone to human error and limits scalability. This study compares various state-of-the-art approaches for requirements identification in an industrial context. For generalizability, we also present an evaluation on a real-world public dataset. [Principal ideas/results] We formulate the requirement identification problem as a binary text classification problem. Various state-of-the-art classifiers based on traditional machine learning, deep learning, and few-shot learning are evaluated for requirements identification based on accuracy, precision, recall, and F1 score. Results from the evaluation show that the transformer-based BERT classifier performs the best, with an average F1 score of 0.82 and 0.87 on industrial and public datasets, respectively. Our results also confirm that few-shot classifiers can achieve comparable results with an average F1 score of 0.76 on significantly lower samples, i.e., only 20% of the data. [Contribution] There is little empirical evidence on the use of large language models and few-shots classifiers for requirements identification. This paper fills this gap by presenting an industrial empirical evaluation of the state-of-the-art approaches for requirements identification in large tender documents. We also provide a running tool and a replication package for further experimentation to support future research in this area.

Autonomous machines are expected to be vastly used at construction sites as they can efficiently perform repetitive and dangerous tasks. However, ensuring the operational safety of such autonomous machines in a highly dynamic environment is challenging. Although autonomous machines usually are equipped with a perception system that permits them to navigate locally, there is a need to share a global view of the construction site to reduce the risk of accidents or errors. A digital twin of the construction site map has the potential of fusing the real-time perception from different sources at the site, such as different autonomous machines working at the construction site, analysing them and sharing the needed information to operate safely and effectively at the site.This paper proposes the adoption of the recently published standard, ISO 23247 digital twin framework for manufacturing, to implement and maintain a dynamic map of construction sites. The proposed framework will enable safe and efficient operation of autonomous machines on construction sites.

Increasing the modal share of the single wagonload transport in Europe requires improving the reliability and predictability of freight trains running between the yards. In this paper, we propose a novel machine learning-assisted macro simulation framework to increase the predictability of yard departures and arrivals. Machine learning is applied through a random forest algorithm to implement a yard departure prediction model. Our yard departure prediction approach is less complex compared to previous yard simulation approaches, and provides an accuracy level of 92% in predictions. Then, departure predictions assist a macro simulation network model (PROTON) to predict arrivals to the succeeding yards. We tested this framework using data from a stretch between two main yards in Sweden; our experiments show that the current framework performs better than the timetable and a basic machine learning arrival prediction model by R2 of 0.48 and a mean absolute error of 35 minutes. Our current results indicate that combination of approaches, including yard and network interactions, can yield competitive results for complex yard arrival time prediction tasks which can assist yard operators and infrastructure managers in yard re-planning processes and yard-network coordination respectively.

Artificial intelligence in manufacturing systems is currently most used for quality control and predictive maintenance. In the lock industry, quality control of final assembled cylinder lock is still done by hand, wearing out the operators’ wrists and introducing subjectivity which negatively affects reliability. Studies have shown that quality control can be automated using machine-learning to analyse torque measurements from the locks. The resulting performance of the approach depends on the dimensionality and size of the training dataset but unfortunately, the process of gathering data can be expensive so the amount collected data should therefore be minimized with respect to an acceptable performance measure. The dimensionality can be reduced with a method called Principal Component Analysis and the training dataset size can be estimated by repeated testing of the algorithms with smaller datasets of different sizes, which then can be used to extrapolate the expected performance for larger datasets. The purpose of this study is to evaluate the state-of-the-art methods to predict and minimize the needed sample size for commonly used machine-learning algorithms to reach an acceptable anomaly detection accuracy using torque measurements from locks. The results show that the learning curve with the best fit to the training data does not always give the best predictions. Instead, performance depends on the amount of data used to create the curve and the particular machine-learning algorithm used. Overall, the exponential and power-law functions gave the most reliable predictions and the use of principal component analysis greatly reduced the learning effort for the machine-learning algorithms. With torque measurements from 50-150 locks, we predicted a detection accuracy of over 95% while the current method of using the human tactile sense gives only 16% accuracy.

Performance aspects such as travel time, punctuality, and robustness are conflicting goals of utmost importance for railway transports. To successfully plan railway traffic, it is therefore important to strike a balance between planned travel times and expected delays. In railway operations research, a lot of attention has been given to construct models and methods to generate robust timetables—that is, timetables with the potential to withstand design errors, incorrect data, and minor everyday disturbances. Despite this, the current state of practice in railway planning is to construct timetables manually, possibly with support of microsimulation for robustness evaluation. This paper aims to narrow the gap between the state-of-the-art optimization-based research approaches and the current state of practice to construct timetables by combining simulation and optimization. The paper proposes a combined simulation-optimization approach for double-track lines, which generalizes previous work to allow full flexibility in the order of trains by including a new and more generic model to predict delays. By utilizing delay data from simulation, the approach can make socioeconomically optimal modifications of a given timetable by minimizing predicted disutility—the weighted sum of scheduled travel time and total predicted delay. In a large simulation experiment on the heavily congested Swedish Western Main Line, it is demonstrated that compared with a real-life, manually constructed timetable, large reductions of delays as well as improvements in punctuality could be obtained for a small cost of marginally longer travel times. The cost of scheduled in-vehicle travel time and mean delay was reduced by 5% on average, representing a large improvement for a highly utilized railway line. Furthermore, a separate scaling experiment indicates that the approach can also be suitable for larger problems. Funding: This research was funded by Trafikverket [Grants TRV 2016/5090 and TRV 2020/72690].

Historically, cylinder locks’ quality has been tested manually by human operators after full assembly. The frequency and the characteristics of the testing procedure for these locks wear the operators’ wrists and lead to varying results of the quality control. The consistency in the quality control is an important factor for the expected lifetime of the locks which is why the industry seeks an automated solution. This study evaluates how consistently the operators can classify a collection of locks, using their tactile sense, compared to a more objective approach, using torque measurements and Machine Learning (ML). These locks were deliberately chosen because they are prone to get inconsistent classifications, which means that there is no ground truth of how to classify them. The ML algorithms were therefore evaluated with two different labeling approaches, one based on the results from the operators, using their tactile sense to classify into ‘working’ or ‘faulty’ locks, and a second approach by letting an unsupervised learner create two clusters of the data which were then labeled by an expert using visual inspection of the torque diagrams. The results show that an ML-solution, trained with the second approach, can classify mechanical anomalies, based on torque data, more consistently compared to operators, using their tactile sense. These findings are a crucial milestone for the further development of a fully automated test procedure that has the potential to increase the reliability of the quality control and remove an injury-prone task from the operators.

Requirements engineering is a time-consuming process, and it can benefit significantly from automated tool support. Ambiguity detection in natural language requirements is a challenging problem in the requirements engineering community. Several Natural Language Processing tools and techniques have been developed to improve and solve the problem of ambiguity detection in natural language requirements. However, there is a lack of empirical evaluation of these tools. We aim to contribute the understanding of the empirical performance of such solutions by evaluating four tools using the dataset of 180 system requirements from the electric train propulsion system provided to us by our industrial partner Alstom. The tools that were selected for this study are Automated Requirements Measurement (ARM), Quality Analyzer for Requirement Specifications (QuARS), REquirements Template Analyzer (RETA), and Requirements Complexity Measurement (RCM). Our analysis showed that selected tools could achieve high recall. Two of them had the recall of 0.85 and 0.98. But they struggled to achieve high precision. The RCM, an in-house developed tool by our industrial partner Alstom, achieved the highest precision in our study of 0.68.

Construction sites are a special kind of off-road environment that needs dedicated dynamic maps to enable autonomous navigation in such terrains. In this paper, challenges for autonomous navigation on construction sites are first identified. Later, requirements for dynamic maps for autonomous navigation on construction sites are proposed based on the identified challenges.

In Sweden and other countries it is not an uncommon practice that freight trains depart more or less on-demand instead of strictly following a pre-planned timetable. However, the systematic effects of freight trains departing late or (in particular) early has long been a contested issue. Although some microscopic simulation tools currently have the capability to evaluate the effect of freight trains departing before schedule, it has yet not been established how macroscopic simulation tools, capable of fast simulation of nation-wide networks, can manage such tasks. This paper uses a case study on a line between two large freight yards in Sweden to investigate how the results of microscopic and macroscopic simulation, represented by two modern simulation tools, differ when it comes to this particular problem. The main findings are that both the microscopic and the macroscopic tools replicated the empirical punctuality fairly well. Furthermore, allowing early departures of freight trains increased overall freight train punctuality while the passenger train punctuality decreased slightly, as determined by both tools. The results are encouraging, but further studies are needed to determine if macroscopic simulation is on-par with microscopic simulation.

Test automation brings the potential to reduce costs and human effort, but several aspects of software testing remain challenging to automate. One such example is automated performance testing to find performance breaking points. Current approaches to tackle automated generation of performance test cases mainly involve using source code or system model analysis or use-case-based techniques. However, source code and system models might not always be available at testing time. On the other hand, if the optimal performance testing policy for the intended objective in a testing process instead could be learned by the testing system, then test automation without advanced performance models could be possible. Furthermore, the learned policy could later be reused for similar software systems under test, thus leading to higher test efficiency. We propose SaFReL, a self-adaptive fuzzy reinforcement learning-based performance testing framework. SaFReL learns the optimal policy to generate performance test cases through an initial learning phase, then reuses it during a transfer learning phase, while keeping the learning running and updating the policy in the long term. Through multiple experiments in a simulated performance testing setup, we demonstrate that our approach generates the target performance test cases for different programs more efficiently than a typical testing process and performs adaptively without access to source code and performance models.

Performance testing with the aim of generating an efficient and effective workload to identify performance issues is challenging. Many of the automated approaches mainly rely on analyzing system models, source code, or extracting the usage pattern of the system during the execution. However, such information and artifacts are not always available. Moreover, all the transactions within a generated workload do not impact the performance of the system the same way, a finely tuned workload could accomplish the test objective in an efficient way. Model-free reinforcement learning is widely used for finding the optimal behavior to accomplish an objective in many decision-making problems without relying on a model of the system. This paper proposes that if the optimal policy (way) for generating test workload to meet a test objective can be learned by a test agent, then efficient test automation would be possible without relying on system models or source code. We present a self-adaptive reinforcement learning-driven load testing agent, RELOAD, that learns the optimal policy for test workload generation and generates an effective workload efficiently to meet the test objective. Once the agent learns the optimal policy, it can reuse the learned policy in subsequent testing activities. Our experiments show that the proposed intelligent load test agent can accomplish the test objective with lower test cost compared to common load testing procedures, and results in higher test efficiency.

Shunting yards are one of the main areas impacting the reliability of rail freight networks, and delayed departures from shunting yards can further also affect the punctuality of mixed-traffic networks. Methods for automatic detection of departures, which are likely to be delayed, can therefore contribute towards increasing the reliability and punctuality of both freight and passenger services. In this paper, we compare the performance of tree-based methods (decision trees and random forests), which have been highly successful in a wide range of generic applications, in classifying the status of (delayed, early, and on-time) departing trains from shunting yards, focusing on the delayed departures as the minority class. We use a total number of 6,243 train connections (representing over 21,000 individual wagon connections) for a one-month period from the Hallsberg yard in Sweden, which is the largest shunting yard in Scandinavia. Considering our dataset, our results show a slight difference between the application of decision trees and random forests in detecting delayed departures as the minority class. To remedy this, enhanced sampling for minority classes is applied by the synthetic minority oversampling technique (SMOTE) to improve detecting and assigning delayed departures. Applying SMOTE improved the sensitivity, precision, and F-measure of delayed departures by 20% for decision trees and by 30% for random forests. Overall, random forests show a relative better performance in detecting all three departure classes before and after applying SMOTE. Although the preliminary results presented in this paper are encouraging, future studies are needed to investigate the computational performance of tree-based algorithms using larger datasets and considering additional predictors.

A quarry site is used to produce gravel with different granularities based on the customer's needs. The production process consists of different stages that need to be planned in order to achieve high productivity and low cost. During the operation of the quarry site, the operating parameters change, and the site productivity could be affected, therefore, a replanning of the production is needed. This paper presents a real-time optimization model for production planning in a quarry site. The optimization model consists of three modules, a productivity monitoring module, a simulation model, and an optimization algorithm. The productivity monitoring module produces the site productivity based on the current operating parameters of the quarry site. Once the site productivity reaches a specific threshold, the optimization algorithm is run on an updated simulation model to find the optimal values of the site parameters. The proposed real-time optimization model is applied to a real quarry site simulated on a high-fidelity simulator. The case study shows that the proposed optimization model can effectively improve the productivity of the quarry site and hence reduces the wastes and improve environmental sustainability.