Making explicit domain knowledge in formal system development

Summary Accurate rate of penetration (ROP) prediction is essential for optimizing drilling performance and reducing operational costs. However, the complexity and uncertainty inherent in drilling operations pose significant challenges to existing data-driven models, limiting their capacity to generalize across diverse conditions and effectively incorporate domain knowledge. In this paper, we propose a knowledge-guided hybrid temporal graph neural network (KHTGNN) for robust ROP prediction. First, explicit domain knowledge is applied to perform physical corrections on drilling parameters during feature engineering, enhancing data quality. Second, an adaptive graph construction method based on node embedding is introduced to dynamically model the evolving correlations among drilling parameters. Third, a hybrid temporal graph neural network (GNN) is designed to capture both intravariable and intervariable dependencies. Moreover, domain-specific rock-breaking mechanism equations are embedded as soft constraints within the model’s loss function to enforce physical consistency. To further enhance generalization, a multiple kernel (MK) maximum mean discrepancy (MMD) approach is integrated to achieve domain-adaptive learning across different wells. The results demonstrate that KHTGNN outperforms current models, achieving a mean absolute percentage error (MAPE) of 9.49%, a mean absolute error (MAE) of 1.64 m/h, and a coefficient of determination (R2) of 0.99. Ablation studies validate the contributions of physical data correction, knowledge constraints, and domain adaptation, emphasizing the critical role of domain knowledge integration in improving predictive performance. This work provides a robust knowledge-guided framework for sequential parameter prediction tasks in complex industrial systems.

We review current and emerging knowledge-informed and brain-inspired cognitive systems for realizing adversarial defenses, eXplainable Artificial Intelligence (XAI), and zero-shot or few-shot learning. Data-driven machine learning models have achieved remarkable performance and demonstrated capabilities surpassing humans in many applications. Yet, their inability to exploit domain knowledge leads to serious performance limitations in practical applications. In particular, deep learning systems are exposed to adversarial attacks, which can trick them into making glaringly incorrect decisions. Moreover, complex data-driven models typically lack interpretability or explainability, i.e., their decisions cannot be understood by human subjects. Furthermore, models are usually trained on standard datasets with a closed-world assumption. Hence, they struggle to generalize to unseen cases during inference in practical open-world environments, thus, raising the zero- or few-shot generalization problem. Although many conventional solutions exist, explicit domain knowledge, brain-inspired neural networks and cognitive architectures offer powerful new dimensions towards alleviating these problems. Prior knowledge is represented in appropriate forms like mathematical relations, logic rules, knowledge graphs, and large language models (LLMs). and incorporated in deep learning frameworks to improve performance. Brain-inspired cognition methods use computational models that mimic the human brain to enhance intelligent behavior in artificial agents and autonomous robots. Ultimately, these models achieve better explainability, higher adversarial robustness and data-efficient learning, and can, in turn, provide insights for cognitive science and neuroscience—that is, to deepen human understanding on how the brain works in general, and how it handles these problems.

An Integrated Intelligent Modeling and Simulation Language for Model-based Systems Engineering

All software systems execute within an environment or context. Reasoning about the correct behavior of such systems is a ternary relation linking the requirements, system and context models. Formal methods are concerned with providing tool (automated) support for the synthesis and analysis of such models. These methods have quite successfully focused on binary relationships, for example: validation of a formal model against an informal one, verification of one formal model against another formal model, generation of code from a design, and generation of tests from requirements. The contexts of the systems in these cases are treated as second-class citizens: in general, the modelling is implicit and usually distributed between the requirements model and the system model. This paper is concerned with the explicit modelling of contexts as first-class citizens and illustrates concepts related to implicit and explicit semantics on an example using the Event B language.KeywordsVerificationmodellingContextsDomains

Appears in: EDULEARN19 Proceedings Publication year: 2019Pages: 1562-1568ISBN: 978-84-09-12031-4ISSN: 2340-1117doi: 10.21125/edulearn.2019.0463Conference name: 11th International Conference on Education and New Learning TechnologiesDates: 1-3 July, 2019Location: Palma, Spain

Critical systems are present in our daily life and affect many aspects providing essential support to several human activities; they are employed in several application domains, providing support to economic activities, transportation, communication, health-care. Such systems must meet strict non-functional requirements, including dependability requirements, and should be able to cope with competitive market needs. Their criticality implies the necessity to meet several requirements, often dictated by international standards, whose fulfillment must be demonstrated in order to achieve necessary certifications. Quantitative evaluations are hence needed to assess and demonstrate compliance with target requirements, since design phases of the lifecycle. For this purpose modeling approaches have been investigated during the years; their aim is to have a of the overall system which allows to assess and demonstrate these properties. It is commonly known that the adoption of formal models in industry, although if strongly recommended when not mandatory, is slowed down by difficulties dictated both by the complexity of the models (which reflects the complexity of the systems) and by the need to have skilled staff in the usage of formal languages. For this reason, also if the academic research trends have focused on complex modeling approaches and techniques, formal models are not so used in industries: in particular only combinatorial models are widely adopted, thanks to their intuitive representation. The great simplifications introduced by combinatorial models lead to approximated results, hence system designers do not trust in them. In the last years the need of a model engineering is noticed: there is the necessity to define appropriate methodologies and processes to support development activities of complex models, aiming at improving the quality/cost ratio. In this way three main research trends have been introduced: formal models generation, multiformalism, and compositional approaches. This thesis aims at defining a methodology for quantitative analysis of critical system that can integrate Model-Based evaluations, conducted by using formal models, with Model-Driven Engineering (MDE) paradigms, taking advantage of the actual research trends. More in detail, according to such approach, critical systems properties can be assessed starting from high level models expressed in proper languages (usually called Domain Specific Modeling Languages - DSMLs), and defining automatic transformation chains, able to generate formal models. The methodology is implementable into cost-effective processes in order to reduce time-to-market specifically during verification stages of critical system development lifecycles. Such objective will be achieved by both improving usability and re-usability of typical Model-Driven artifacts. The adoption of this methodology for railway system dependability analysis may lead to an estimated costs reduction of approximately 30%. The usage of a high-level central language significantly decreases also the training time of new analysts, as they only need to learn the DSML usage (properly developed for the application domain). This thesis defines also two high-level modeling languages: the first, CIP_VAM, has been defined to vulnerability aspects of physical infrastructures, while the second, DAM-Rail, allows the dependability modeling in railway domain. Another important contribution of this thesis is given in the definition of new methods and formal operators: in detail a reference architecture for performability and Quality of Service evaluation is defined, as well as the formal concept of Model Template is introduced (to enable reuse). At last two techniques which deals with scalability are addressed: stub and reduced models. Even if the focus is on quantitative evaluation, the proposed methodology is useful also for the complete design (based on centralized informations), hence it is possible to extend it for automatic generation of documents and, more generally, of all those artifacts which support system development. The proposed approach has been applied with different goals to case studies coming from three different domains: railway systems, conferencing systems, and sensor network. These systems have different and complementary properties in terms of number of involved components, scalability, etc. Produced models are described, and obtained results are shown and commented.

As biological sequences and various annotation data grow rapidly in public databases, the classification problems become larger and more complicated. New classifier designs are necessitated. Besides, how to incorporate some explicit domain knowledge into learning methods is also a big issue. In this paper, we adopt a modular classifier, min-max modular support vector machine (M3-SVM) to solve protein subcellular localization problem, and use the domain knowledge of taxonomy information to guide the task decomposition. Experimental results show that M3-SVM can maintain the overall accuracy and improve location average accuracy compared with traditional SVMs. The taxonomy decomposition is superior to other decomposition methods on a majority of the classes. The results also demonstrate a speedup on training time of M3-SVM compared with traditional SVMs.

The chapter presents a description of a novel modelling language called Alvis defined for the design of concurrent especially real-time systems. Alvis combines the advantages of formal methods and practical modelling languages. Based on CCS and XCCS process algebras, Alvis utilizes flexible graphical modelling of interconnections among agents and a high level programming language used for the description of agents behaviour. Each agent in a model can use rule-based systems to support its decisions. A small set of language statements and graphical concepts make Alvis easy to learn and use. The possibility of a formal model verification, makes Alvis a formal modelling language. Alvis modelling environment creates in parallel a model of the considered system and a labelled transition system (LTS graph) that is its formal representation. The LTS graph can be formally verified with the CADP toolbox. A survey of main Alvis features from practical point of view, is given in the chapter.

Clinician scheduling remains a persistent challenge due to limited clinical resources and fluctuating demands. This complexity is especially acute in large academic anesthesiology departments as physicians balance responsibilities across multiple clinical sites with conflicting priorities. Further, scheduling must account for individuals' clinical and lifestyle preferences to ensure job satisfaction and well-being. Traditional approaches, often based on statistical or rule-based optimization models, rely on structured data and explicit domain knowledge. However, these methods often overlook unstructured information, e.g., free-text notes from routinely administered clinician well-being surveys and scheduling platforms. These notes may reveal implicit constraints (post-call recovery needs, early departures for family obligations) and underutilized clinical resources (a physician's willingness to work additional hours to address emergent needs). Neglecting such information can lead to misaligned schedules, increased burnout, overlooked staffing flexibility, and suboptimal utilization of available resources. To address this gap, we propose a predict-then-optimize framework that integrates classificationbased clinician availability predictions with a mixed-integer programming schedule optimization model. Large language models (LLMs) are employed to extract actionable preferences and implicit constraints from unstructured schedule notes, enhancing the reliability of availability predictions. These predictions then inform the schedule optimization considering four objectives: (i) ensuring clinical full-time equivalent compliance, (ii) reducing workload imbalances by enforcing equitable proportions of shift types, (iii) maximizing clinician availability for assigned shifts, and (iv) schedule consistency. By combining the interpretive power of LLMs with the rigor of mathematical optimization, our framework provides a robust, data-driven solution that enhances operational efficiency while supporting equity and clinician well-being.

Towards automatic building extraction from high-resolution digital elevation models

Systems architecting has been considered for a long time both an art and a science. The systems architect uses heuristics, stories, and models to communicate complex architectural concepts to stakeholders. As adoption of model-based systems engineering grows, and digital models of system architecture increase in use, there is a danger that the communications gap between system architects and the stakeholders they are working for will grow. Previous work created a competency model exploring the link between art, systems thinking, and systems architecture. A further conference panel entitled "If architectures are so useful, why are they so seldom used?" inspired a discussion of the growing communications gap between the system architect as modeler and the system architect as communicator. Decision makers cannot read and interpret formal system models. However, principles and practices of systems architecture are exhibited in the creation of film scores, fine arts, and building architecture. Reflecting on the authors' experience as practitioners and educators of systems thinking and systems architecting, this paper discusses the need for formal education in the arts as a way to bridge the communication problems that technical architects have with their stakeholders.

Railway vehicles are generally maintained preventively within certain time periods. Condition based predictive maintenance strategies have a great economic potential so that modern trains are equipped with many sensors in order to perform diagnostics and prognostics of components.Methods for fault detection need appropriate feature subsets in order to achieve small in-sample and out-sample errors. In our case the typical feature selection approach using pure data-driven methods is difficult, as the number of possible feature sets is very large. On the other hand there exists rich domain knowledge and detailed physical models of the mechanical system. The aim is to combine this knowledge with the often used mathematical methods for feature selection for improving classification of cases when a faulty damper is present. Based on the dynamic equations of motion, this paper presents heuristic feature selection via the analysis of transfer functions. We describe several wellknown methods of automated feature selection and a workflow which combines domain knowledge with automated methods. Results show that it is difficult to definefeatures based only on domain-knowledge, but in combination with data-driven techniques good classification performance can be achieved.

We attempt in this paper to outline a method for the automatic construction of a knowledge base. We propose some methods and a domain knowledge model. A new idea is to conceive a system that is able to each phase of its construction to acquire domain knowledge from all new information that it is building, in particular the indexing terms; the last section is an attempt in this sense.

We describe the use of the Cyber-Physical Modeling Language (CyPhyML) to support trade studies and integration activities in system-level vehicle designs. CyPhyML captures integration interfaces across multiple design domains for system components, and generic design assembly rules given in terms of architecture alternatives. The CyPhyML tools support automated exploration of system-level architectural and parametric tradeoffs using a suite of design exploration tools that can be applied to models at different levels of fidelity and scale. Our overall approach includes exploration over the space of potential designs by evaluating structural combinations and then comparison of designs by simulating the dynamics of systems and subsystems with varying degrees of detail. In that flow, we use the DESERT toolkit for design tradeoffs that can be evaluated from the structure of the design model (i.e. interconnections of components and their parameters). DESERT extends a graphical modeling language with concepts and relations to define structural alternatives and constraints on system properties. Alternatives are given in an abstract way, decoupled from the details of the encoding of the combinatorial design space problem for the underlying binary decision diagram (BDD) solver. For desirable design instance models, the tools automatically assemble complete vehicle or subsystem computer-aided design (CAD) models from the associated component CAD models, and likewise create detailed finite-element structural analysis (FEA) models for selected component assemblies. Evaluation results are presented in a dashboard display which provides comparisons between different valid designs over all specified design metrics.

The paper discusses the method for C4ISR capability requirement analysis. The method focuses on domain knowledge modeling and domain knowledge reusing. The approach uses three layers framework to describe capability requirement, which makes the domain knowledge reusable. Based on the framework, the process for C4ISR capability requirement modeling has been discussed. The meta-concept model has been discussed in the paper too. Finally, a case study of C4ISR architectural simulation modeling is provided to demonstrate the availability and applicability of the method.