http://2015.semantics.cc/tudor-b-ionescu

1. Architecting Semantic Process Mining-Driven
Optimization and Adaptation of Business
Workflows for the Mobility Industry
Tudor B. Ionescu
Software Architect @SIEMENS

2. Safety Critical Applications  Processes are also safety critical!
SIEMENS Rail Automation Applications

3. Business Processes in Rail Automation
• Complex business processes described in natural language
– Conventional documents (MS Word, PDF)
– Process execution logs
– Emails, meeting minutes, ...
• Business process =
Unstructured data + Tacit knowledge + Workflows
• Safety, reliability, certifiability, and standard compliance
 Key to the success of products and projects
• Fulfilling these non-functional requirements is extremely
costly and time-consuming
 Automation and optimization desired
3

4. Agenda
• Introduction
– Mission/safety-critical business processes in the railway domain
• Challenges of Process Automation and Optimization
• A Three-Process Model for Automation and Optimization
• Implementation in Camunda
• Process Recovery
• Conclusion
4

5. Challenges of Process Automation and
Optimization
• Extracting semantic process models from unstructured data
– handbooks, logbooks, execution logs of tools and past projects
• Enacting the automation and optimization of business
processes according to the mined semantic models
• Seamlessly and dynamically adapting running processes
whenever
– Unexpected potentially harmful situations occur
– New insights are gained by means of process mining
– New safety compliance requirements become available
• Architecting reliable process architectures for mission/safety-
critical applications
5

6. Example Process
• Siemens Mobility
– Process for rail automation plant
• CIRCE
• AUTOCAD
• Excel, CSV, PDFs
• File Lists, signal tables, …
• Email and other communication means
• Legacy technologies (SAMBA, X25 …)
• People
• Standards & Constraints  Verification
– Goal
• Automate & optimize (some of) these tasks
– Success criteria
• The automated process is really used
• The automated process makes life easier not harder
1) Service Tasks =
Facades for external
resources
2) Notifications from
external services
and actors (e.g.,
emails)
3) Which patterns
(e.g., publish-
subscribe, polling)
4) Interfaces (REST or
legacy?)
5) Data flow
6) Deployment
7) Process recovery
6

7. Task
Approach: A Three-Process Model
Task
Task
Adapted
task
Adapt
Suspend and
modify
Process
Input &
Output
Task
Require
+ Design
+ Decide
Inform
Security Problem
Optimization
potential
Business
Process(es)
Mining
Process(es)
Monitor
+ Learn
Adaptation
Process(es)
KB
7

8. Process Mining and Adaptation
• Mining process
– Runs in parallel with the productive business processes
– Its main task consists of mining the unstructured process data
– Results (insights) are given in
• Natural language  Expert-driven adaptation (for safety and/or mission
critical processes)
• RDF format (JSON-LD)  Automated adaptation (for non-critical
processes)
– Insights are used to inform the adaptation process
• Adaptation process
– Uses insights to optimize and adapt the main process
– Adaptation is realized without interrupting the main process
– Must ensure recovery in case of failures caused by adaptations
8

9. Example insight
• Mining task:
– Mining GIT (versioning system) logs produced by the process
• Result: Reverse engineered Gantt chart (Ref…)
– BALA, S., CABANILLAS, C., MENDLING, J., ROGGE-SOLTI, A., POLLERES, A. (2015). Mining
Project-Oriented Business Processes. Lecture Notes in Computer Science 9253, 425-440.
• Insight: Process can be optimized  How?  See Gantt Chart
• Requirement: Adapt the processes accordingly
9

10. Implementation: Camunda Suite
• Pros
+ Addresses all aspects of process management & execution
+ High-quality documentation
+ Sound software architecture (high design pattern density)
+ Uses state of the art technology (JAX-RS, AngularJS, Spring, jQuery, etc.)
+ Many supported environments (Tomcat, JBOSS, Glassfish …)
+ Flexibility, extensibility (Community extensions)
• Cons?
- Is it really lightweight?
• A constellation of tools  Temptation to use them all  Complexity
- Infrastructure needs  Web servers, database servers 
Administrators + Programmers
- Real flexibility or exhaustiveness?
• Bottom line
+ Pros seem to dominate cons from a software architecture perspective
10

11. The Adaptation Process
• Uses insights to optimize and adapt the main process
• Adaptation is realized without interrupting the main process
• Must ensure recovery in case of failures caused by adaptations
11

12. What about recovery in case of failures?
• Ad hoc software/process updates may introduce latent
software faults
– Obvious errors are corrected early (e.g. during development and
testing)  subtler errors remain in the code and are more difficult to
detect
– The more complex the process  the harder it is to make it reliable
and safe
– There is a finite amount of ressources to be spent on testing
 Safety/mission-critical processes must have means for
recovery at runtime (switch to reliable process)
• Idea  Use forward recovery mechanism
– Sha, L. (2001). Using simplicity to control complexity. IEEE Software,
(4), 20-28.
12

13. Forward Recovery for Safety/Mission-
Critical Processes
• Challenge
– The root of software faults is complexity
• But: cost reduction and optimization can only be achieved through
process adaptation, which may lead to more complex processes
• Design Tactic
– Use a simple and reliable core process that provides the essential
service
– Decompose the process in adaptable subprocesses
– Adapt one subprocess at a time  version_1 (core), version_2
(adapted)
– Keep core version in stand-by in case adapted version fails
• Prerequisite: You need a simple and reliable core process
– But: if you can‘t develop a simple and reliable process  there are
chances you will not be able to develop an optimized reliable process
13

14. Subprocess Adaptation Example
Task automation
14

15. Forward Recovery
• Detect faulty behavior using an acceptance test or safe output
region check
• Keep/derive an analytic relationship between the outputs of
the core and adapted versions of the subprocesses, e.g.
– DIFF: the modified process should only produce an output that is
measurable and differentiable with respect to the old version of the
process
• Camunda @runtime subprocess deployment mechanism
15

16. Demo
• saFiddle
– Web-based tool for editing and managing requirements and design
decisions  Integrates into the adaptation process
– Uses concept linkage between architectural elements and concepts
(e.g. requirements, quality attributes, design patterns and tactics, etc.)
16

17. Conclusion
• Process adaptation and optimization is a desideratum 
Reduces costs
• In safety/mission-critical contexts
– Adaptation & optimization cannot be automated
– Non-functional requirements such as safety and reliability must be
fulfilled at all times
– Means for process recovery must be in place
• The three-process model is able to fulfill these requirements
• The integration into the adaptation process of web-based
tools for
– Designing safe and reliable adaptations of subprocesses
– Creating traceable requirements and design decisions
– Linking faults and errors to design decisions  Enable a learning
process
17

18. Acknowledgements
• This work has been funded by the Austrian Research
Promotion Agency (FFG) under grant 845638 (SHAPE).
18