This paper presents an architecture for automated fault recovery in special-purpose machinery using standard IEC 61131-3 PLC control software. Building upon the MAPE-K paradigm (Monitor, Analyze, Plan, Execute with Knowledge), the proposed three-level architecture integrates runtime control via a previously presented modularization concept called Control Primitives, high-level reasoning using Regionalized Resource Process Dependence Graphs (RRPDGs), and an intermediate Resource Agent that bridges field-level execution and recovery planning. The architecture supports fully automated dynamic recovery and seamless resumption of productive operations while maintaining real-time reliability. Experimental evaluation demonstrates the scalability and low resource demands of the planning system across simulated scenarios and validates the concept on a lab-sized real-world demonstrator. The architecture and interfaces build a modular foundation for future research and industrial applications.

Automated Production Systems (aPS) in regulated industries such as pharmaceuticals, MedTech, or food and beverages must comply with the stringent validation and documentation requirements of Good Manufacturing Practice (GMP) regulations within the European Union (EU). These obligations create significant burdens for aPS manufacturers, particularly when changes require revalidation through manual integration and system testing. But these prerequisites also enable opportunities for lower-effort software verification, leveraging GMP documentation and boundaries of the automation domain to prove, e.g., that an implemented software change realizes the change specification without side effects. This paper proposes a methodical workflow for deriving software slices suitable for formal verification, while being aligned with automation engineering practices and GMP requirements. It defines assumptions for slice utility based on system modularity, interface expressiveness, and domain boundaries. The approach is validated through a real-world GMP-regulated change from a German MedTech aPS manufacturer, following GAMP 5 guidelines, demonstrating the utility for GMP-regulated aPS engineering and automatic verification of a sliced program segment.

In automated production, resilience describes a system’s capacity to absorb disturbances by reconfiguring itself, thus retaining its Overall Equipment Effectiveness at least partially. This includes online behavior reconfiguration to automatically recover from or prevent faults, collectively called fault handling. Promising research exists for fault handling in automated Production Systems. In process engineering, fault diagnosis and automatic parameter adaptions are already industrially available. However, handling faults in discrete manufacturing requires a series of distinct operations, which cannot be achieved by parameter changes alone. Further, core requirements must remain fulfilled by automatic fault handling approaches, including real-time control and extra-functional requirements like changing operation modes, monitoring interlocks, and an alarming and communication system. This article proposes a concept for reconfigurable IEC 61131–3 software for automatic fault handling, validated by a public reference implementation for a demonstrator, an industrial production system, and a modified industrial test rig. Eight experiments were successfully conducted, showcasing four use cases of the concept: The prevention of faults by avoiding anomalous components, the recovery from a fault state to automatic operation, the definition of previously undefined state variables, and the monitoring of global interlocks to trigger a controlled stop. All mentioned extra-functional requirements are fulfilled. Note to Practitioners—Identification, reporting, diagnosis, and recovery of faults in automated production incur substantial effort. Project-specific code is required for diagnosis, and the recovery and re-initialization are often performed manually. To our knowledge, automatic recovery approaches from scientific literature are not widely used in discrete manufacturing. Reasons may include a frequent disregard of extra-functional requirements mentioned above. Further, some approaches are incompatible with IEC 61131–3 or industry-typical software modularization. This article proposes a PLC software concept that aims to be compatible with real-world challenges and solutions. The functional software is vertically modularized from organizational hardware-level code. The horizontal modularization separates devices or equipment groups. Support for multiple changing operation modes including two types of controlled stop (run to completion or abort), alarming, data exchange, and global interlocks are incorporated. A prototypical IEC 61131–3 implementation is publicly available that separates a reusable generic part from hardware-specific and project-specific code. The resulting control code is highly reusable, such that all modes (derived from PackML), including dynamic reconfigurations, are composed from the same software modules. Note that we do not expect the concept to be well-adoptable in continuous processes, as elaborated in the Preliminaries section.

The design of manufacturing systems requires exploring diverse component configurations for one that best satisfies the requirements of the target system. At the same time, manufacturing systems evolve continuously, increasing their complexity and that of the production line in which they are installed. Model-based System Engineering (MBSE) methodologies and modeling languages like System Modeling Language (SysML) are used to assist system engineers during the entire design process. Yet, MBSE methodologies mainly rely on manually composed models for requirements verification, incurring enormous cost and effort especially in larger, more complex systems. This paper proposes a methodology exploiting the knowledge enclosed in a SysML model to automate the execution of a set of analyses defined over the represented manufacturing system. The proposed approach relies on Lingua Franca (LF) to simulate the specified analysis and analyze the results. Moreover, the methodology supports chains of analyses to represent dependent requirements and tradeoffs imposed upon a manufacturing system. The proposed methodology has been evaluated on a manufacturing system demonstrator. Results show the proposed methodology's applicability and scalability on diverse configurations of the manufacturing system.

Industrial alarm systems facilitate the reliable and effective management of industrial process operations. However, the manufacturer’s fear of missing a critical fault or lacking process understanding may lead to the implementation of alarm systems that overburden operators by an enormous amount of alarms. Inadequate alarm settings and maintenance, coupled with a high volume of disturbance alarms, necessitate operators making crucial decisions within a limited time. Thus, an alarm management strategy that accurately anticipates the types of incoming alarms is needed, especially in brownfield systems where it is infeasible to simplify the underlying alarming functions. This alleviates the need for human interpretation of alarms upon arrival, allowing operators to address anomalous behaviors early. This study proposes an alarm classification approach, based on active learning, latent dirichlet allocation (LDA) topic modeling and transformer-based deep learning. It focuses on addressing these challenges through a systematic natural language processing (NLP) approach to classify and map fault messages. To demonstrate the validity of this approach, real-world alarm data from an industrial hybrid non-woven manufacturing process is used.

Faults in the technical process of automated production systems (aPS) are challenging to detect, report, and recover accurately. Missing physical equipment, lacking detection mechanisms in the control software, and insufficient training and experience of human operators lead to an often delayed and imprecise detection of a fault's root cause. Since fault detection and recovery typically rely heavily on human involvement, there is also a risk of operators performing only partial repairs or inadvertently introducing new faults. Thereby, even slight deviations in human-machine interactions may affect process outcomes unpredictably, which so far cannot be modelled in field-level automation. The Functional Resonance Analysis Method (FRAM) is well-suited to address functional relations in such complex socio-technical systems and accurately represent resource requirements, preconditions, control, and time constraints, making it a promising tool to methodically investigate aPS faults. Yet, the FRAM has barely been applied in this domain. This paper thus introduces an approach relying on the FRAM to methodically derive alarm conditions and system extensions as a foundation for broader uses to investigate in future work. The FRAM-based approach was shown to reduce uncaught faults, required operator interventions, and thus unplanned downtime, in an experiment with two participants on a lab-sized demonstrator machine.

The importance of high-performance computers in car networks increases to realize software assistance functions. These computers offer new opportunities for use cases in car testing and diagnostics towards vehicle software and services. Therefore, car testing and diagnostics are no longer limited to the testing and functional checking of electric and electronic components. Proven concepts from computer science, such as business process management, can be integrated into the car network to provide a uniform basis for automotive use cases. In this paper, (i) the Business Process Model and Notation (BPMN) standard is used to model diagnostic processes in future car production lines and (ii) a car-tailored BPMN workflow engine is introduced to execute the processes in a high-performance vehicular computer. The research on executable BPMN models in car networks bridges the gap between interdisciplinary business processes and execution in automotive IT systems.

Automated production systems (aPS) are variant-rich, design-to-order systems and an increasing proportion of their functionality is implemented by control software. In control software development, software reuse is still commonly performed via clone-and-own despite many drawbacks, e.g., copying errors. This unplanned reuse leads to a high amount of historically grown software variants, which contain valuable domain expertise. Therefore, to enable planned reuse of existing control software solutions, an analysis of legacy software, inducing documentation of identified variability, is required. While so-called Software Product Lines enable the documentation of variability, they lack suitable variability visualization tailored to the needs of aPS stakeholders such as application or module developers. To address this gap, this paper introduces a variability visualization concept tailored to the needs of aPS stakeholders with the aim of supporting them in their daily tasks. The concept was evaluated successfully within a master student's course by use of a prototypical implementation of the visualization concept.

The development of automated Production Systems (aPS) is an interdisciplinary process, where an increasing part of the system's functionality is realized in the respective control software. Such software projects commonly utilize programming languages standardized in IEC 61131-3. To measure, improve, and maintain source code while also promoting trust in its capabilities, an objective assessment of its characteristics is necessary. Software metrics are a means for such an evaluation. While there is an abundance of metrics available from the classical software engineering domain, these metrics focus on textual programming languages. IEC 61131-3, however, defines graphical languages, which are not targeted by renowned concepts in computer science. Besides, former research demonstrates that software engineering metrics for textual languages need adaption to be applicable in the aPS domain. Thus, this paper introduces a metrics suite consisting of adapted and newly developed measures, which focus on the graphical IEC 61131-3 language Function Block Diagram. The results are prototypically implemented in one of the leading integrated development environments for IEC 61131-3 and then evaluated regarding their understandability and applicability by practitioners at a German aPS manufacturer.