Analysis of industrial processes Process mining is a method for identifying, analyzing and improving business processes using data. The approach is already established in certain industries such as online commerce, but not yet in industrial processes. The reasons: Processes very rarely have large amounts of data (Big Data), which is a prerequisite for process mining methods, and there is a lack of the appropriate corporate structures and methods to implement such technology. The goal of the project 'BPM-I4.0' is the holistic development, implementation and evaluation of process mining methods for the analysis and prescriptive control of industrial core processes. To this end, innovative procedures, concepts, algorithms and digital tools are prototypically designed, implemented, evaluated, processed and generalized in the product creation and order fulfillment processes of participating companies. The results achieved will enable companies to significantly improve the quality of their core processes based on the analysis of their process data and to proactively control process execution in order to maintain and expand their competitiveness in the medium and long term.

Problem How can process mining be applied in knowledge-intensive industrial processes?

With regard to the implementation of process mining (PM) in industrial core processes, there are essentially three major challenges:

1) The interdisciplinary nature of PM with elements from IT, BPM and Data Science often prevents a clear technical integration into the organizational structure. It is therefore all the more important to define a dedicated corporate function (so-called BPM Office), which has the necessary skills and resources to make PM methods usable in the company.

2) PM tools and methods must be adapted or further developed according to the specialist context of industrial companies. 

3) As such, industrial core processes (e.g. in order processing) represent a particular challenge for PM methods. They are often knowledge-intensive processes, with many customer- and product-specific variants. These characteristics are often associated with a lack of data, so additional data sources have to be found.

Development of methods and tools for the use of process mining in knowledge-intensive processes

The project uses the methods of classical process mining and combines them with different machine learning methods based on knowledge-intensive processes (Figure 1). Approaches were explored to help companies get started with process mining as well as to make process mining feasible for application to knowledge-intensive processes. Through the results, companies should gain the ability to further explore the topic of process mining and establish this for other processes.



Following the definition of the use cases and the verification whether they are suitable for process mining, the focus is on an assistance system that should enable companies to make better decisions in the processes. Furthermore, a specification technique for the description of data sources, IT systems and expert information or domain knowledge in the process context must be developed. It forms the basis for the Machine Learning (ML) models. Furthermore, the development of a process mining maturity model was another focus. The goal here is to develop a framework for action for successful business transformation in the dimensions of people, organization, technology starting from a maturity classification to prescriptive process mining. In total, a procedure model for the implementation of process mining is created here, which aims in particular to improve the ability to optimize their own business processes.



A set of tools was developed, which will serve companies in the future as a practical tool to facilitate the entry with process mining, insb. for knowledge-intensive processes. The developed tools and methods were evaluated during the project based on real use cases. Exemplary for industrial core processes are here the product development process (Weidmüller) and the order processing process (GEA). Thus, a range of new methods, concepts and tools have been developed in the project, which support the use of process mining for knowledge-intensive processes as well as recommendations for action in the process.

.

Successful application of process mining methods in knowledge-intensive use cases

In summary, the project goals have been comprehensively achieved to a large extent. The project has produced a data canvas for process mining projects. The Process Mining Data Canvas (PMDC) enables users to navigate through early phases of process mining projects in a structured manner. In addition, the PMDC enables project participants to identify data problems and increase data quality.

Furthermore, a KPI system was developed in conjunction with a method for designing KPI profiles. This way, goals and their measurability can be defined in the context of the process mining project at the same time.

A comprehensive maturity model was developed, which allows companies to determine their maturity with regard to the use of process mining. Whether SME or global player, the model enables each company equally to assess and progressively build the technical and organizational capabilities required for PM. In addition to the holistic approach of a maturity model, an online survey (quick check) has been developed through which companies can get a first overview of their individual maturity level.

Finally, a prototype for prescriptive process mining has also been implemented. With the help of different ML approaches, future delays in process flows can be estimated, for example. Based on these predictions, which were implemented by means of predictive process mining in the project, the prototype makes use of the principle of Diverse Counterfactual Explanations (DICE). So that alternative courses of action can be generated that, for example, prevent delays in a process.