Safety-critical Human- and dAta-centric Process management in Engineering projects
Large-scale deployments of technical infrastructure products such as industrial plants, railway automation or power stations, which themselves involve complex engineering processes, are a crucial part in the value-creation chain of production systems for large-scale infrastructure providers.
The following challenges demanding enhanced ICT support in this domain have been identified from real cases at Siemens:
Challenge 1: Integrated description of processes, constraints, resources and data.
The availability of an integrated description of processes is of paramount importance for efficient and effective status monitoring, constraint and consistency verification and checking of compliance rules. It also provides the basis for managing resource constraints and availability. Up until now, various formal languages are at hand for describing processes, resources and data separately, but no integrating model exists that supports such rich querying functionality.
Challenge 2: Integration and monitoring of structured and unstructured data.
To a high degree, engineering steps are the input for state changes of the process. Often, these state changes are only implicitly visible as manipulation of data. Therefore, an en- gineering process must also incorporate these data smoothly for monitoring control flow, version updates, data storage, and email notification. To this end, various types of systems have to be integrated including their structured and unstructured data (e.g. by mail traffic, or ticketing systems).
Up until now, these data need to be integrated mostly manually and with delay.
Challenge 3: Documentation of safety-critical aspects.
Each engineering project has its time-critical phases, where engineers are under heavy load. Especially, these phases are prone to sloppy process documentation and reduced quality of results. The BPMS should not generate extra work which could be done automatically.
Notably, many of the steps are required to be documented in prescribed ways by safety and security standards and regulations (such as SIL). Semi-automatic mapping of witnessing data to fulfillment of these security regulations would significantly help in compliance checking with such regulations.
Challenge 4: Be ready for changes.
Monitoring all the engineering activities and applying appropriate process patterns may lead to modification proposals of the process to make it simpler and less error-prone. Notably, many of the steps are required to be documented in prescribed ways by safety and security standards and regulations (such as SIL). Semi-automatic mapping of witnessing data to fulfillment of these security regulations would significantly help in compliance checking with such regulations.
Currently, adaptations require manual work and are prone to errors.
Challenge 5: Acceptance and human factors.
The overall process management needs to be set up in a non-obtrusive way, such that engineers working the processes will find it useful and easy to use. This is a specific challenge in safety-critical systems, which are developed with a tight timeline. It calls for a design that integrates existing tools and working styles instead of introducing new systems.
The overall goal of the project is to leverage ICT support for more rigorous and verifiable process management in recurring and adaptive engineering processes occurring in large-scale industry and infrastructure projects. In order to address the challenges mentioned above, the following goals were identified:
Goal 1: We will develop an integrated semantic model to describe and monitor processes, resources, constraints and data, i.e. devise a unified querying language and algorithms to monitor and run processes. Our model shall be integrated with existing process engines, but allows for more rigorous monitoring of resource and data flow constraints at the process instance level. This addresses Challenge 1 by means of enabling more rigorous support within existing ICT systems for safety-critical processes.
Goal 2: To address Challenge 2, we need to devise mechanisms to detect and extract structured from unstructured process data in the day-to-day engineering work witnessing processes. To this end, we will leverage methods to integrate structured and unstructured data, gather structured process models from unstructured informal process descriptions by deploying methods from Natural Language Processing (NLP) and Ontology learning.
Goal 3: In order to address Challenge 3, we aim to instantiate our semantic models in the concrete domain of safety-critical process engineering tasks (including models of SIL).
Goal 4: In order to address Challenge 4, we need to harness our methods and algorithms to be flexible, adaptive and robust. Thus we aim at developing methods to deal with ad-hoc activities and changed processes (detect ad-hoc activities, reconfigure resources according to changed processes).
Goal 5: Finally, we aim to prototypically deploy and validate the results in a real-world use case from the railway domain, in which we will conduct empirical user evaluations.
In the SHAPE project we will propose ICT support for more rigorous and verifiable process management in such recurring and adaptive engineering processes. In particular, we will:
Develop a combined semantic model for processes, resources and compliance rules, drawing from Business Process Modeling Languages and declarative knowledge representation & reasoning (such as BPMN, SPARQL, Answer Set Programming) that allows verifiable processes and declarative modeling of constraints.
Extend and devise specific methods for detecting and extracting structured processes from unstructured processes and data (such as text extraction and process mining).
Develop domain-specific semantic models in the concrete of safety-critical process engineering tasks (such as SIL).
Enrich the above methods to be robust and adaptive through re-planning and re-configuration features.
Devise an architecture and validate the results in a real-world use case from the railway domain, in which we will conduct empirical user evaluations.
An architectural blueprint and the distribution of work in work packages (WP) is depicted in the next figure: