Keywords

These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.

1 Problem Statement

The Industry 4.0 vision aims at creating Smart Factories by combining the Internet of Things, Internet of Services, and Cyber-Physical Systems. To support this vision, Industry 4.0 communities have fostered the definition of standards such as AutomationML (IEC 62424) and OPC UA (IEC 62541). AutomationML is one of the core standards of Industry 4.0 for exchanging plant engineering information as specified by [1, 7, 16, 18]. AutomationML can describe plant components and their sub-components from different views such as mechanical, electrical, or software. OPC UA [4] also allows for the description of the production life-cycle in Smart Factories, but contrary to AutomationML which describes characteristics of plant components, OPC UA models machine-to-machine communication. Although Industry 4.0 standards provide the basis for data exchange in a Smart Factory, information models of these standards require being aligned to facilitate the merging of a virtual process with real production life-cycles.

Smart Factories along with Cyber-Physical concepts impose new challenges to traditional approaches of data integration. A new generation of data-centric systems need to be developed and integrated, where data meaning, as well as data variety, veracity, and adaptivity must be managed [12]. In this context, achieving semantic data interoperability techniques suitable for these data properties, is of paramount importance for making the Industry 4.0 vision a reality.

With the aim of assessing the integration of Industry 4.0 standards, Biffl et al. [1] and Kovalenko and Euzenat [9] have characterized seven semantic heterogeneity issues among different views of an artifact. (M1) Value processing–same properties are not modeled equally, e.g., using different datatypes; (M2) Granularity–same objects are modeled at different levels of detail; (M3) Schematic differences–differences in the way how semantics is represented for the same object; (M4) Conditional mappings–relations between entities exist only if certain conditions occur; (M5) Bidirectional mappings–relations between entities have to be defined bidirectionally; (M6) Grouping and aggregation–different semantic modeling criteria are applied to group elements for the same object; and (M7) Restrictions on values–mandatory values for properties in the object that have to be handled in the mapping process. These semantic heterogeneity issues may occur between documents defined in the same or different standards, i.e., interoperability can be intra- or inter-standard.

Industry 4.0 standards and initiatives for identifying heterogeneity issues evidence the success of the Industry 4.0 movements. However, integration of standards are still conducted manually [5, 17], negatively affecting the effectiveness of the production process. This doctoral work attempts to achieve two research goals to solve these problems. RG1: Addressing intra-standard interoperability among multiple pieces described in one standard, e.g., AutomationML. RG2: Assessing inter-standard interoperability in documents specified in different standards, e.g., documents in AutomationML and OPC UA.

To accomplish our research goals RG1 and RG2, we propose a rule-based system that combines Deductive databases and Semantic Web technologies to effectively integrate documents specified in Industry 4.0 Standards.

2 State of the Art

Related work is divided into two sub-sections: (1) approaches for solving intra-standard interoperability with the focus in AutomationML; and (2) approaches for addressing inter-standard interoperability on AutomationML and OPC UA.

Intra-standard interoperability issues for AutomationML. In the literature, many different approaches are proposed for integrating AutomationML documents. In [17], a tool to map two AutomationML files is presented. It allows for the integration of AutomationML documents, their respective descriptions, and the modified parts of one file into the other. Further, a mapping algorithm for AutomationML files is presented. Nevertheless, the process of mapping is performed manually. Himmler [8] presents a framework to create standardized application interfaces in plant engineering based on AutomationML. The work provides a function-based based standardization framework for the plant engineering domain. Persson et al. [13] utilize an RDF-based approach to integrate robotized production information modeled AutomationML. Kovalenko et al. [10] explore how AutomationML can be represented by means of Model-Driven Engineering and the Semantic Web. A small part of an AutomationML ontology is developed, based on the main concepts of the language. Also, the use of rules for consistency checking is proposed, using the Semantic Web Rule Language (SWRL), but no explicit definition of the role of Semantic Web technologies on the integration problem is presented. The AutomationML Analyzer [14] is an on-line tool to browse, query and analyse different AutomationML data by means of Semantic Web technologies. A conceptual design to overcome integration problems in AutomationML is described. All these approaches have the potential to solve specific integration problems for AutomationML. However, they solve rather isolated problems, and a general method capable to automatically integrate AutomationML information from different perspectives is not provided.

Inter-standard interoperability issues for AutomationML and OPC UA. The current integration approach is performed by analyzing common information elements and manually describing the AutomationML objects in the OPC UA language [5, 7, 15]. In addition, in these works are documented the mappings between the common objects, i.e., their common structures and datatypes. To date, the possibility to semi-automatically integrate the information models of these two languages exploiting their semantic descriptions is still missing.

3 Proposed Approach

We propose Alligator [6], a rule-based framework for the semantic integration of Industry 4.0 standard documents. Alligator relies on Datalog and RDF to accurately represent the knowledge that characterizes different types of semantic heterogeneity for these documents. Further, by utilizing Datalog as well as Semantic Technologies, Alligator will be able to provide explanations for the alignments that occurred among the elements of the Industry 4.0 standard documents.

Figure 1 depicts the main components of the proposed solution. In the following we describe each component. The input for Alligator are Industry 4.0 standard documents. Next, these documents are translated into a canonical representation of a production life-cycle named the Alligator data model. RDF, RDFS, and domain-specific vocabularies are used to represent the core concepts of the Alligator data model. Further, standards documents are modeled as facts in an extensional database (EDB) of a Datalog program. Datalog rules comprise an intensional database (IDB), and state different types of semantic heterogeneity and intra- and inter- standard interoperability problems. The Alligator Deductive System Engine performs a bottom-up evaluation of Alligator Datalog programs following a semi-naïve algorithm that stops when the least fixed-point is reached [2]. The intensional predicates inferred in the evaluation of a Alligator Datalog program correspond to problems of semantic heterogeneity, and intra- and inter-standard interoperability. AutomationML, which is an XML-based standard, is translated into RDF using Krextor [11], an XSLT-based framework for converting XML to RDF. In addition, the mapping rules for the conversion using Krextor have to be created according to the AutomationML vocabulary, and an RDFS vocabularyFootnote 1 and the mapping rulesFootnote 2 for the AutomationML standard have been defined. Similarly, we plan to define more rules to cover other standards and to semantically integrate them.

Fig. 1.
figure 1

The Alligator Architecture. Alligator receives documents in different Industry 4.0. standards, and creates an integrated document. Input documents are represented as RDF graphs and Datalog predicates (EDB) in the Alligator data model. Datalog intentional rules (IDB) characterize semantic heterogeneity types. A bottom-up evaluation of the Datalog program identifies semantic heterogeneity inconsistencies among input documents

Full size image

We present an example of one of the designed Datalog rules that states when two AutomationML elements are the same. Based on the AutomationML specification, this condition is met depending on links with external standards such as eCl@ss [3], which semantically identify elements. Accordingly, if two elements contains the same eClassIRDI as a value, then they are semantically equivalent. Listing 1.1 is a Datalog rule that represents this knowledge.

figure a

4 Results

Testbeds. With the aim of testing the effectiveness of Alligator, we generated 30 Testbeds for intra-standard interoperability issues focusing in AutomationML. Testbeds are based on the semantic heterogeneity types M2 (granularity), M3 (schematic differences), and M6 (grouping and aggregation); ten testbeds per each type of heterogeneity. First, a seed (AutomationML document) was manually created for each testbed according to the type of semantic mapping. Next, we automatically generated two AutomationML documents derived from this seed containing a random number of semantic equivalent AutomationML elementsFootnote 3. The generation was performed following a uniform distribution. Testbeds corresponded to pairs of AutomationML documents, and thirty testbeds were evaluated in the studyFootnote 4. Further, a Gold Standard was manually generated computing the elements that were semantically equivalent as well as those different ones. For the compilation of a Gold Standard, we relied on the generated testbeds.

Table 1. Effectiveness of Alligator. Per semantic heterogeneity type, effectiveness of Alligator is reported. In all testbeds, precision is 1.0. Alligator exhibits the highest performance in the testbeds of type M2 (F-measure is always 1.0), while in M3 and M6, the F-measure values are at least 0.8.
Full size table

Table 1 reports on the values of these metrics for each type of semantic heterogeneity, i.e., M2, M3, and M6. We observed that for these semantic heterogeneity types, the value for precision is 1.0, i.e., Alligator correctly detected all the semantically equivalent elements. Further, recall and F-measure are also 1.0 in the testbeds of semantic heterogeneity M2. These results suggest that Alligator rules capture the knowledge required to accurately solve the AutomationML semantic equivalent elements Identification problem. For the semantic heterogeneity types M3 and M6, Alligator rules are not completely covering all possible semantic equivalences generated between nested structures of AutomationML elements. Thus, Alligator could not identify at most two semantic equivalent elements in five out of 20 testbeds of type M3 and M6.

5 Methodology

The methodology adopted in this doctoral work comprises the following tasks:

  1. 1.

    Investigation of state-of-the-art approaches relevant to the problem of integrating Industry 4.0 standards.

  2. 2.

    Formalization of the problem of integrating Industry 4.0 standards and proposed solutions; definition of research questions and hypotheses of our formal and empirical study are stated. Alligator is designed under the following hypothesis: an approach combining Datalog rules and Semantic Web technologies, for the integration of Industry 4.0 standards will exhibit better performance than the state of the art approaches. In addition, we identified the following research questions: (RQ1) Is Alligator able to identify pairs of semantic equivalent elements in Industry 4.0 documents? (RQ2) Does Alligator exhibit equal behavior whenever different types of semantic heterogeneity occur during the integration of Industry 4.0 documents?

  3. 3.

    Empirical evaluation of our hypothesis to measure Alligator performance with respect to state-of-the-art approaches.

    1. (a)

      Implementation of state-of-the-art or baselines approaches.

    2. (b)

      Definition of benchmarks to evaluate the proposed solutions.

    3. (c)

      Design and execution of experiments and statistical tests to validate or falsify our hypotheses. The result section was a first attempt to measure the effectiveness of Alligator. We designed a controlled experiment where some of the heterogeneity types were measured. Based on this idea we plan to extend the Testbeds to cover the following:

      1. i.

        All the heterogeneity types, i.e., from M1 to M7 for the intra- as well as for the inter-standard interoperability issues.

      2. ii.

        A validation with domain experts to evaluate results of the integrated documents, as well as the generated explanations. This validation will be conducted for both problems, intra- and inter- standard interoperability issues, and for all the heterogeneity types.

    4. (d)

      Analysis and discussion of the observed results.

6 Conclusions and Future Work

The problem of assessing intra- and inter-standard interoperability is addressed in this work, and we propose the combination of Deductive databases and Semantic Web technologies to effectively solve these problems. As a proof-of-concept, we present Alligator, a deductive framework for the integration of Industry 4.0 Standard documents. Alligator relies on Datalog and RDF to accurately represent the knowledge characterizing different types of semantic heterogeneity for documents described in Industry 4.0 standards. Currently, the main focus of our approach is to solve semantic heterogeneity issues that may occur among documents defined in the same, i.e., intra-standard interoperability problem. Nevertheless, we additionally plan to extend Alligator to integrate different standards and assess the inter-standard interoperability problem.

The results of the empirical evaluation for the intra-standard interoperability problem using AutomationML standard, suggest that Alligator is able to effectively solve the problems of AutomationML semantic equivalent Identification element and exhibits similar behavior for the three studied semantic heterogeneity types, i.e., granularity (M2), schematic (M3), and grouping (M6). In the future, we will empower the Alligator Deductive System Engine with the expressiveness of Datalog with negation and built-in predicates. Thus, Alligator will be able to represent all the other types of semantic heterogeneity in AutomationML, i.e., from M1 to M7. The heterogeneity types will be also defined for the inter-interoperability problem, i.e., between AutomationML and OPC UA. Further, we will extend Alligator to create explanations of the aligned elements of the integrated Industry 4.0 standard documents. Finally, we envision to develop a more general framework, capable of semantically integrate documents expressed in different Industry 4.0 standards.