Heiko Koziolek, Andreas Burger, Hadil Abukwaik, Julius Rückert, Marie Platenius-Mohr Industrial Automation, Corporate Research Ladenburg, Germany, email@example.com, firstname.lastname@example.org, email@example.com, julius.rückert@de.abb.com, firstname.lastname@example.org
Creating control logic and process graphics for control and monitoring systems such as ABB Extended Automation System 800xA, ABB Ability™ Symphony Plus, or ABB Freelance still requires a significant amount of manual work. While engineering libraries provide low-level reuse and bulk engineering tools automate the handling of I/O lists, automation engineers must still translate many customer specifications by hand. Consequently, the DEXPI industry initiative – supported by BASF, Equinor and Bayer – is currently working on standardized plant topology models to automate additional engineering tasks . An ABB research project demonstrated how these models could save costs and increase quality. The findings are based on the analysis of four plant specifications from recently executed System 800xA projects.
Plant engineering today
Today’s process engineers specify piping and instrumentation diagrams (P&IDs) as blueprints for the automation of a production process →01. These drawings describe the required equipment – such as tanks, pumps, motors, and valves – and instrumentation, eg, sensors for temperature, flow, level and pressure. Although there are naming conventions and industry standards for the shapes of the components in P&IDs, different computer-aided design (CAD) tools offer process engineers a lot of freedom. They can create custom shapes, add free-text annotations and unintentionally draw unconnected pipes – all of which complicates algorithmic analysis and prevents automated processing of the encoded information.
Consequently, P&IDs are often exchanged as PDF files (or even printouts) from which control logic and process graphics are manually derived by automation engineers. Based on their experience, the engineers can compensate for some semantical ambiguities (eg, shapes that do not fully conform with standards). However, they also often encounter inconsistency or incompleteness in the diagrams in relation to other specification documents, which leads to time-consuming communication feedback loops with the process engineers.
A few CAD tools already feature support for so-called Smart P&IDs, but these are not yet widely used in the industry. Smart P&IDs include a database that contains structured tables of the encoded information (eg, instrumentation lists) and metadata for the drawn items (eg, pipe diameter, alarm limits, etc.). Algorithms can process such structured information much more easily than drawings consisting of generic boxes, lines and circles. Currently, Smart P&IDs are usually stored in formats specific to a particular CAD tool, which complicates the construction of software tool chains.
Since 2011, the DEXPI initiative has worked on a common P&ID specification standard, expressed as object-oriented concepts in an XML file format. This standard can be considered to be a standardized version of Smart P&IDs. The specification captures both drawing information (eg, graphical coordinates and drawing instructions) and abstract models of equipment, instruments and their dependencies. The latter are also called “topology models” since they express the plant topology as a kind of network, analogous to the topologies of electronic circuits or computer networks . The initiative is driven by large automation customers, such as BASF, Bayer, Covestro, Equinor, Evonik and Merck. All major CAD tool vendors are also involved, for example, Autodesk, Aveva, Hexagon and Siemens →02. The specification has matured in recent years (Version 1.2 was released in 2020) and CAD tool vendors frequently participate in DEXPI working group hackathons (intense software development sessions) to test the DEXPI XML importers and exporters that will find their way into the next product release.
Topology models are at the center of “topology engineering,” developed over the past few years in scientific communities, which aims at utilizing topological information to carry out engineering tasks. Standardized models in DEXPI XML file format allow tasks that are currently performed manually to be automated:
• Control logic generation. To some extent, interlocking logic and state-based control can be derived from plant topology models .
• Process graphics generation. The graphical layout derived from a P&ID and encoded in a topology model can serve as a template for the creation of human-machine interfaces for plant operators.
• Simulation generation. Topology models can be mapped to object types in simulation frameworks (eg, Modelica), thereby creating low-fidelity plant simulators to use in factory acceptance tests and to train operators.
• Root-cause analysis. The operator can query plant topology models to investigate the root causes of anomalies .
• Alarm management. Cascades of alarm messages in an industrial plant may overload human operators; plant topology models can be used to limit such “alarm floods.”
CAYENNE Topology Editor
A recent ABB research project has implemented a prototype software tool called “CAYENNE Topology Editor” to demonstrate the possible automation of engineering tasks using plant topology models. The tool supports the creation of plant topology models from DEXPI XML files, from Microsoft Visio P&IDs and from proprietary SmartPlant P&ID exports →03. In brownfield projects, topology models can also be derived from existing 800xA process graphics, which contain coarse-grained topological information. Users can inspect and edit imported topology models visually.
The CAYENNE tool provides a control-logic generator that can synthesize interlocking logic from the topology models. A rule engine supports the generator and applies predefined, domain-specific rules to topology models in order to generate the control logic. For example, if a level indicator on a tank issues a “low” alarm, a pump on an outlet of this tank is stopped. The rule engine traverses the topology model and searches for the pattern encoded in the rule. Once a match is found, it retrieves the relevant tag names and generates the required control logic, conditionally linking the alarm condition signal for the cause to the control signal relating to the effect. The tool supports the generation of System 800xA Control Builder M control diagrams and function block diagrams. In addition, IEC 61131-11 Structured Text can be generated, as well as cause-and-effect matrices.
CAYENNE process graphics generator
Also provided by the CAYENNE tool is a process graphics generator, which maps the layout imported from a P&ID to the shapes contained in an ABB System 800xA engineering library. This enables the partial generation of System 800xA process graphics, which can then be completed manually by an automation engineer. The tool supports generating equipment, instruments and pipes. The positions and sizes of the shapes are preserved, which involves translating between the graphical coordinate system of the topology model and the process graphic. This mapping can be customized for different System 800xA engineering libraries so that their specific shapes can be displayed in the process graphics. This procedure has been demonstrated for the System 800xA “standard” library and the “reuse” library that is normally used for oil and gas plants.
Integration with AUCOTEC’s Engineering Base
The CAYENNE Topology Editor has been integrated with AUCOTEC’s Engineering Base tool – used by ABB in greenfield projects – as a prototype. AUCOTEC is currently implementing a DEXPI XML importer for Engineering Base, which will enable ABB’s Plant Data Processing (PDP) Tool to be combined with a topology model. The CAYENNE Topology Editor can generate process graphics from Engineering Base as well as create interlocking logic that “glues” together the function blocks generated by the tool.
To evaluate topology engineering and to investigate if the CAYENNE Topology Editor could have sped up engineering, ABB conducted four retrospective studies on specifications from automation projects in plants already erected and automated :
• A mid-sized fertilizer production plant in South America with around 1,000 I/Os →04. The evaluated plant segment contained 18 vessels, eight pumps and a reactor.
• A fuel production plant (4,000 I/Os) in South America.
• An oil separation process (400 I/Os) in the Middle East that featured separation vessels, instrumentation and a sophisticated piping structure.
• An upstream oil production process (7,000 I/Os) in South America →05. This process included a large number of parallel, similar pipelines and instrumentation, so only a selection was analyzed.
For each case, a research team analyzed the EPC plant specifications and selected a representative plant segment, which consisted of 10 to 20 P&IDs per case. Afterward, topology models for the CAYENNE Topology Editor were created. As the P&IDs were only available as PDF files, this work required the redrawing of the P&IDs in formats supported by the CAYENNE tooling. For example, the research team created Microsoft Visio P&IDs based on the PDF files, which took roughly one day per case. Once EPCs export their P&ID files in the DEXPI XML file format, this step can be omitted. Importing the created files into the CAYENNE Topology editor yielded the required topology model.
Afterward, the research team analyzed around 50 to 100 interlocks per case, specified in cause-and-effect matrices. For each cause/effect pair, the connecting process topology path within the P&ID was looked up. By comparing this part to the path of other cause/effect pairs, generic interlock concepts were identified and encoded as rules. For example, a commonly observed pattern was that if a pressure sensor raised a “high” alarm, a preceding valve needed to be closed. For many interlocks the generic rule needed to be defined only once and could later be applied multiple times. If these rules had been available before the project, the cause-and-effect matrices could have been generated to a large extent using the CAYENNE Topology Editor.
In total, 91 percent of the interlocks in the case studies could be generated by rules. The case studies yielded 92 interlocking rules in total, 73 percent of which were classified as “generic,” meaning that they would likely apply in other plants. The remaining 27 percent of the rules contained plant-specific patterns that may hold only for very similar plants; the effort expended on these is only justified if they occur numerous times within the plant.
A few (7 percent) of the rules were applicable across all four case studies. The reason for this low percentage is the plants’ heterogeneity: Where, for example, fertilizer production mainly concerns level measurements, the upstream oil production almost exclusively concerns pressure measurements in pipelines. Selecting cases with greater similarity (eg, five fertilizer plants) would likely increase the cross-case reuse of interlocking rules significantly.
Estimated cost saved by interlock generation and the elimination of manual coding and testing is around 15 percent of the overall effort for control logic engineering. Human error sources are also removed. Based on the case studies, the estimated cost savings for process graphics generation are even higher, at around 50 percent.
A significant step toward automating engineering
Topology models are key elements for automation of engineering tasks for ABB’s System 800xA, ABB Symphony Plus Operations and ABB Freelance. All major vendors of CAD tools are working on support for the recently developed DEXPI XML standard for PIDs. Topology models extracted from these diagrams enable partial control logic and process graphics generation and have been used to generate plant simulators for operator training in the past. To reap the benefits of topology models, more case studies need to be conducted to improve tooling and concepts. The required software tooling will be optimized for usability and integrated into other engineering tools. This step will take the automation industry a long way down the road to fully automated engineering.
 E. Arroyo et al., “From paper to digital,” ABB Review 1/2016, pp. 65–69.
 Schleburg M. et al., “A combined analysis of plant connectivity and alarm logs to reduce the number of alerts in an automation system,” Journal of Process Control, 23(6), pp. 839–851, 2013.
 Drath R. et al., “Computer-aided design and implementation of interlock control code,” 2006 IEEE Conference on Computer Aided Control System Design, 2006 IEEE International Conference on Control Applications, 2006 IEEE International Symposium on Intelligent Control, pp. 2,653–2,658, October 2006.