Distributed control systems (DCS) are essential to operate complex industrial processes. A major part of a DCS is the alarm system, which helps plant operators to keep the processes stable and safe. Alarms are defined as threshold values on individual signals taking into account minimum reaction time of the human operator. In reality, however, alarms are often noisy and overwhelming, and thus can be easily overlooked by the operators. Early alarm prediction can give the operator more time to react and introduce corrective actions to avoid downtime and negative impact on human safety and the environment. In this context, we introduce Alarm Prediction Transformer (APT), a multimodal Transformer-based machine learning model for early alarm prediction based on the combination of recent events and signal data. Specifically, we propose two novel fusion strategies and three methods of label encoding with various levels of granularity. Given a window of several minutes of event logs and signal data, our model predicts whether an alarm is going to be triggered after a few minutes and, if yes, it also predicts its location. Our experiments on two novel real industrial plant data sets and a simulated data set show that the model is capable of predicting alarms with the given horizon and that our proposed fusion technique combining inputs from different modalities, i. e. events and signals, yields more accurate results than any of the modalities alone or conventional fusion techniques. • Multimodal Transformers can successfully predict alarms in industrial settings. • Transformers can predict future alarms based on recent events and signal data. • The proposed hybrid fusion performs on par with or better than early or late fusion. • Granular alarm encoding yields more accurate prediction than the whole tag.

Machine learning (ML) has become popular in the industrial sector as it helps to improve operations, increase efficiency, and reduce costs. However, deploying and managing ML models in production environments can be complex. This is where Machine Learning Operations (MLOps) comes in. MLOps aims to facilitate this deployment and management process. One of the MLOps challenges is understanding how ML models reason, which is key to trust and acceptance. Here, explainable AI (XAI) can help. Better error identification and improved model accuracy are only two resulting advantages. An often neglected fact is that deployed models are bypassed when model performance or explanations do not meet user expectations. In this paper, we provide a novel reference architecture to address the challenge of integrating explanations and feedback capabilities into MLOps. Our architecture is implemented in a series of industrial use cases in the project EXPLAIN. The proposed MLOps software architecture has several advantages. It provides an efficient way to manage ML models in production environments. Further, it allows for integrating explanations into the development and deployment processes.

With the increasing application of artificial intelligence (AI) and machine learning (ML), the topic of technical debt management for machine learning systems is gaining more attention. Additionally, industrial systems including manufacturing or logistics processes are also supposed to benefit from AI and ML, which is reported in many publications related to ML application models. However, fewer studies on “how is technical debt managed in context of ML systems” are being published. This contribution fills this gap by reporting findings from 15 semi-structured and in-depth interviews conducted with industrial practitioners. Based on the interview results, suggestions for an initial technical debt management process and two document artifacts that facilitate the process are addressed.

Classification with multivariate signal data is an important machine learning task in artificial intelligence applications in the process industry. Examples of such applications range from process monitoring and optimization, product quality prediction, or predictive maintenance. Signal data captures physical quantities like pressures, flows, levels, temperatures, vibrations, etc. Although relative large historical data sets are available in process plants, a common problem in the development of classification models - especially for process monitoring and optimization - is the lack of labels. We introduce a first version of an active learning web-application that can support human experts in providing labels for the identification of steps in batch recipe from process data while gaining insights into the learning progress of the machine learning model.

Autonomy is today a widely discussed topic. Despite the already impressive technical possibilities, autonomy will still happen stepwise in industry. When it comes to paving the way toward autonomy in industry, it is important to define an appropriate taxonomy of autonomy levels from an industry perspective. Based on such a basic taxonomy, this paper suggests a process how to apply it to specific industrial systems. Exemplarily, an application to autonomous plants is sketched. Besides, enabling technologies required toward autonomy are discussed, as well as some implications of autonomy in industry.

Data preparation for data mining in industrial applications is a key success factor which requires considerable repeated efforts. Although the required activities need to be repeated in very similar fashion across many projects, details of their implementation differ and require both application understanding and experience. As a result, data preparation is done by data mining experts with a strong domain background and a good understanding of the characteristics of the data to be analyzed. Experts with these profiles usually have an engineering background and no strong expertise in distributed programming or big data technology. Unfortunately, the amount of data can be so large that distributed algorithms are required to allow for inspection of results and iteration of preparation steps. This contribution introduces an interactive data preparation workflow for signal data from chemical plants enabling domain experts without background in distributed computing and extensive programming experience to leverage the power of big data technologies.

For supporting interpretation, assessment and application of data mining models, explanation-aware methods are crucial. This paper presents an approach for explanation-aware feature selection and assessment using symbolic abstractions of time series. For that, we utilize the symbolic approximate aggregation (SAX) method for data abstraction to be implemented into data mining models. We investigate several approaches and discuss experiences in the context of petro-chemical production.

With the advent of cloud computing and its model for IT services based on the Internet and big data centers, the interest of industries into XaaS ("Anything as a Service") paradigm is increasing. Business intelligence and knowledge discovery services are typical services that companies tend to externalize on the cloud, due to their data intensive nature and the algorithms complexity. What is appealing for a company is to rely on external expertise and infrastructure to compute the analytical results and models which are required by the business analysts for understanding the business phenomena under observation. Although it is advantageous to achieve sophisticated analysis there exist several serious privacy issues in this paradigm. In this paper we investigate through an industrial use-case the application of a framework for privacypreserving outsourcing of pattern mining on event-log data. Moreover, we present and discuss some ideas about possible extensions.

Big Data technologies enable new possibilities to analyze historical data generated by process plants. One possible application is the development of new types of operator support systems (OSS), which could help plant operators during operations in identifying and dealing with critical situations. The project FEE has the objective to develop such support functions based on Big Data analytics of historical plant data. In this contribution we describe our approach to define software architectures for Big Data enabled OSS in industrial plants.

Dispersed data sources, incompatible data formats and a lack of non-ambiguous and machine readable meta-data are major obstacles in data analytics and data mining projects in process industries. Usually, meta-information is only available in unstructured format optimized for human consumption. This contribution captures common problems when handling data from process plants in analytics, develops the vision of a data collection process supported by a search tool, and describe a search tool for the heterogeneous plant data as a proof-of-concept.

Industrial applications from areas like automation, process control, or power controls have a very long-life time up-to 30 years or even more. Supporting applications developed for such life-times cannot rely on the availability of the hardware the application was originally developed on. Especially FPGA families are updated approximately every 12-18th months. This presents a major challenge if functionality in such long-living applications has to be realized on FPGAs (e.g. for performance reasons): How to avoid considerable re-development efforts when replacing obsolete hardware? Concepts for developing portable FPGA applications that can be easily migrated to new hardware are of crucial importance to answer this challenge. In this paper we review and evaluate different approaches for developing portable FPGA applications.

Optimizing the Quality-of-Service (QoS) levels of a service workflow is essential for the user satisfaction in Service-oriented Computing. For that purpose, QoS computation models are applied to reflect the actual QoS experienced by the user during service execution. Current QoS models ignore the possible dependencies of QoS attributes, such as the dependency on the time of the execution or on the input data supplied to the service. Apart from that, composition approaches consider only single workflows during service selection, narrowing the number of possible compositions. Thus, we introduce a novel QoS model that covers QoS dependencies and discuss how this model can be used to consider multiple workflows at the same time. Moreover, we adopt a multi-objective optimization approach to offer solutions varying in QoS such as finishing time and price, allowing the user to make fine-grained decisions.

New approaches in the design of manufacturing systems, such as adaptive and service-oriented manufacturing systems or self-optimizing resources introduce new degrees of freedom into manufacturing control. Fast adaptation to quickly changing market requirements and new objectives summarized by the term sustainability are the main drivers of these new concepts. The resulting research question is how human decision makers can exploit the high adaptability of these new manufacturing systems in accordance to the current requirements arising from the market. This contribution reviews a scheduling model suitable to meet the requirements of the new class of manufacturing systems, called evolving manufacturing systems. To solve scheduling problems defined by this model a multiobjective evolutionary approach is suggested and a suitable problem encoding and corresponding evolutionary operators are introduced. To consider the operative character of scheduling, a practicable parallelization strategy suitable for modern CPU designs is introduced.

Web service composition provides a means of customized and flexible integration of service functionalities. Quality-of-Service (QoS) optimization algorithms select services in order to adapt workflows to the non-functional requirements of the user. With increasing number of services in a workflow, previous approaches fail to achieve a sufficient reliability. Moreover, expensive ad-hoc replanning is required to deal with service failures. The major problem with such sequential application of planning and replanning is that it ignores the potential costs during the initial planning and they consequently are hidden from the decision maker. Our basic idea to overcome this substantial problem is to compute a QoS optimized selection of service clusters that includes a sufficient number of backup services for each service employed. To support the human decision maker in the service selection task, our approach considers the possible repair costs directly in the initial composition. On the basis of a multi-objective approach and using a suitable service selection interface, the decision maker can select compositions in line with his/her personal risk preferences.

Self-optimizing mechatronic systems are a new class of technical system promising new levels of flexibility and utility in electro-mechanical systems. Planning is an important method to realize self-optimization, although today hardly used in mechatronics. In this context, planning is understood as search for a feasible sequence of operations which implements the execution of specific job assigned to a system. This search is a complex and time-consuming task. Hence, it is desirable to decompose the planning task into smaller sub problems according to paradigm of divide & conquer and use problem specific solution methods. Unfortunately, possible planning sub problems in mechatronic systems cannot be considered isolated since sub modules influence each other. This paper introduces the application of a multi-agent-planning model based on cooperative objective functions that enable the coordinated solution of sub problems.

Adaptive and service-oriented manufacturing proposes manufacturing systems which can be rapidly changed in accordance to changing production programs and customer individual production. Increasing customer specific manufacturing and new trends such as Sustainable Manufacturing create a complex system of objectives. Human dispatcher controlling these manufacturing systems must be properly supported to consider all required objectives. Due to the reduced insight into the adapting manufacturing systems, classical single optimization approaches with a rigid definition of a single objective function and hard optimization constraints are not appropriate any more. This paper introduces a decision making framework based on multi-objective scheduling. The decision making process consists of a multi-objective scheduling process and an interface for schedule selection. The interface combines the decision making approaches outranking and preference elicitation.

Self-optimizing mechatronic systems are a new class of technical systems. On the one hand, new challenges regarding dependability arise from their additional complexity and adaptivity. On the other hand, their abilities enable new concepts and methods to improve the dependability of mechatronic systems. This paper introduces a multi-level dependability concept for self-optimizing mechatronic systems and shows how planning can be used to improve the availability and reliability of systems in the operating stages.

The joined travelling of vehicles is an important instrument of cost reduction in the innovative RailCab concept. While the problem of joining groups of vehicles into convoys and determining convoy routes can be easily understood as optimization problem, the distributed nature and large number of vehicles and stops inhibits the direct application of operations research methods and problems. In this paper we introduce a combination of multiagent planning techniques like distributed matchmaking and filtering, data clustering from computational intelligence and heuristics from operations research to solve the complex task of convoy formation in the RailCab system. It is shown how the solution space of the optimization problem can be efficiently reduced during the matchmaking by applying filtering mechanisms and clustering to identify groups of agents with compatible optimization constraints

Nature inspired methods are a popular mean to control multiagent system (MAS) in logistic scenarios. Methods like pheromones or potential fields are used as primary control mechanism or for job assignment. Although empty runs are one of the main cost driver in logistics, this aspect is rarely considered in the job assignment procedures in logistic MAS. In this paper we introduce a combination of the pheromones and potential fields to estimate the attractivenes of transport jobs. The attractivenes will be estimated depending on two aspects: the chance to acquire a follow-up job and the waiting times of costumers. The mechanism is introduced in the context of an innovative railway concept called RailCab. The systems does not rely on central control instance and can be entirely implemented by peer to peer communication.

Hardware failures in autonomous and distributed software systems create the need for self-healing activities. This work addresses the problem of redeploying software components affected by a hardware failure, while respecting several runtime constraints. In contrast to existing approaches, we do not only compute a new feasible system configuration, but make use of AI planning in order to derive a sequence of concrete deployment and undeployment actions that achieve this state. We also distinguish between application of every intermediate plan found and the single application of the final plan. These two plan application strategies are described and evaluated along with our novel solution approach.