IMPROVE - Innovative Modelling Approaches for Production Systems to Raise Validatable Efficiency

The integration of smart devices into the production process results in the emergence of cyber-physical production systems (CPPSs) that are a key part of Industrie 4.0. Various sensors, actuators, Programmable Logic Controllers (PLCs), Manufacturing Execution Systems (MES) and Enterprise Resource Planning (ERP) systems produce huge amounts of data and meta data that can hardly be handled by conventional analytic methods. The main goal of this work is to develop an innovative architecture for handling big data from various heterogeneous sources within an automated production system (aPS). Moreover, enabling data analysis to gain a better understanding of the whole process, spotting possible defects in advance and increasing the overall equipment effectiveness (OEE), is in focus. This new architecture vertically connects the production lines to the analysts by using a generic data format for dealing with various types of data. The presented model is applied prototypically to a lab-scale production unit. Based on a message broker, the presented prototype is able to process messages from different sources, using e.g. OPC UA and MQTT protocols, storing them in a database and providing them for live-analysis. Furthermore, data can be anonymized, depending on granted access rights, and can be provided to external analyzers. The prototypical implementation of the architecture is able to operate in a heterogeneous environment supporting many platforms. The prototype is stress tested with different workloads showing hardly any response in the form of longer delivery times. Thus, feasibility of the architecture and its suitability for industrial, near real-time applications can be shown on a labscale.

With this paper, we will illustrate the synergetic potential of interdisciplinary research by demonstrating how socio-scientific perspectives can serve engineering purposes and contribute to engineering assignments. This especially concerns eliciting knowledge models and ergonomically optimizing technological design. For this purpose, we report on our findings on socio-technical arrangements within smart factory research initiatives that are part of the IMPROVE project. We focus our findings on the systemic interplay between the formally modelled plant, its actual physical state and the social environment. We also look at how operators, as parts of the plant’s environment, adapt themselves and thereby develop their own particular work culture. We then integrate these findings by reconstructing this operator work culture as a specific way of performing, accounting for and addressing particular issues. We enhance these findings using the lenses of recent concepts developed in the field of social theory, namely the praxeological understanding of tacit knowledge, systems theory differentiations and an actor-network-theory understanding of human-machine agency. Applying these concepts from social theory, we revisit our empirical findings and integrate them to provide context-sensitive, socioscientifically informed suggestions for engineering research on knowledge models concerning HMI design.

Model-based diagnosis is a commonly used approach to identify anomalies and root causes within cyber-physical production systems (CPPS) through the use of models, which are often times manually created by experts. However, manual modelling takes a lot of effort and is not suitable for today’s fast-changing systems. Today, the large amount of sensor data provided by modern plants enables data-driven solutions where models are learned from the systems data, significantly reducing the manual modelling efforts. This enables tasks such as condition monitoring where anomalies are detected automatically, giving operators the chance to restore the plant to a working state before production losses occur. The choice of the model depends on a couple of factors, one of which is the type of the available signals. Modern CPPS are usually hybrid systems containing both binary and real-valued signals. Hybrid timed automata are one type of model which separate the systems behaviour into different modes through discrete events which are for example created from binary signals of the plant or through real-valued signal thresholds, defined by experts. However, binary signals or expert knowledge to generate the much needed discrete events are not always available from the plant and automata cannot be learned. The unsupervised, non-parametric approach presented and evaluated in this paper uses self-organizing maps and watershed transformations to allow the use of hybrid timed automata on data where learning of automata was not possible before. Furthermore, the results of the algorithm are tested on several data sets.

Modern Cyber-Physical Production Systems provide large amounts of data such as sensor and control signals or configuration parameters. The available data enables unsupervised, data-driven solutions for model-based anomaly detection and anomaly localization: models which represent the normal behavior of the system are learned from data. Then, live data from the system can be compared to the predictions of the model to detect anomalies and perform anomaly localization. In this paper we use self-organizing maps for the aforementioned tasks and evaluate the presented methods on real-world systems.

In the present work, fault detection in industrial automation processes is investigated. A fault detection method for observable process variables is extended for application cases, where the observations of process variables are noisy. The principle of this method consists in building a probability distribution model and evaluating the likelihood of observations under that model. The probability distribution model is based on a hybrid automaton which takes into account several system modes, i.e. phases with continuous system behaviour. Transitions between the modes are attributed to discrete control events such as on/off signals. The discrete event system composed of system modes and transitions is modeled as a finite state machine. Continuous process behaviour in the particular system modes is modeled with stochastic state space models, which incorporate neural networks. Fault detection is accomplished by evaluation of the underlying probability distribution model with a particle filter. In doing so both the hybrid system model and a linear observation model for noisy observations are taken into account. Experimental results show superior fault detection performance compared to the baseline method for observable process variables. The runtime of the proposed fault detection method has been significantly reduced by parallel implementation on a GPU.

The aim of industrial alarm flood analysis is to assist plant operators who face large amounts of alarms, referred to as alarm floods, in their daily work. Many methods used to this end involve some sort of a similarity measure to detect similar alarm sequences. However, multiple similarity measures exist and it is not clear which one is best suited for alarm analysis. In this paper, we perform an analysis of the behaviour of the similarity measures and attempt to validate the results in a semi-formalised way. To do that, we employ synthetically generated floods, based on assumption that synthetic floods that are generated as ’similar’ to the original floods should receive similarity scores close to the original floods. Consequently, synthetic floods generated as ’not-similar’ to the original floods are expected to receive different similarity scores. Validation of similarity measures is performed by comparing the result of clustering the original and synthetic alarm floods. This comparison is performed with standard clustering validation measures and application-specific measures.

In view of the increasing amount of information in the form of alarms, messages or also acoustic signals, the operators of systems are exposed to more workload and stress than ever before.We develop a concept for the reduction of alarm floods in industrial plants, in order to prevent the operators from being overwhelmed by this flood of information. The concept is based on two phases. On the one hand, a learning phase in which a causal model is learned and on the other hand an operating phase in which, with the help of the causal model, the root cause of the alarm sequence is diagnosed. For the causal model, a Bayesian network is used which maps the interrelations between the alarms. Based on this causal model the root cause of an alarm flood can be determined using inference. This not only helps the operator at work, but also increases the safety and speed of the repair. Additionally it saves money and reduces outage time. We implement, describe and evaluate the approach using a demonstrator of a manufacturing plant in the SmartFactoryOWL.