Business process simulation enables testing and improving redesigned versions of business processes in a secure environment. By utilizing process simulators, the efects (e.g., the resource utilization) of process changes can be studied prior to real-world implementation. With the growing availability of process execution data, techniques have emerged to extract process simulation models from event logs. This enables a faster, more evidence-driven approach instead of manually creating simulation models. However, event logs may contain noise and lack specific information, such as resource availability, for a comprehensive simulation model. This demo presents a parameterization tool for simulation models that allows users to improve discovered models by human input or external data and develop diferent redesign scenarios - a user-friendly interface between existing discovery techniques and process simulators.
Organizations run business processes to deliver products and services successfully. Regular
changes in the business process environment, such as increased demand or the availability of
new technology, require improving and adapting the business processes frequently [
SimuBridge 1) Extract and improve as-is simulation scenarios 2) Create new, manage and compare scenarios 3) Start simulation runs
the real world, new process simulation scenarios can be created to test the impact of 1) possible changes in the process environment, e.g., an increase in customers, or 2) new redesign ideas, e.g., changing the order of certain activities or the automation of activities.
As more and more execution data of business processes is available in IT systems, it becomes
increasingly feasible to apply process mining to the discovery of process simulation models [
In this demo, we present SimuBridge, a tool that connects simulation discovery, human input, and process simulators and ofers management of process scenarios and their comparison (cf. Figure 1). In the remaining sections, we present the tool features and architecture, then give insights into application use cases and future developments.
In the following, we first describe the main features of SimuBridge and then its technical structure. Its source code repository, including also a screencast and a tutorial, can be found at https://github.com/INSM-TUM/SimuBridge.
As shown in Figure 1, SimuBridge supports several features explained in this subsection.
1) Extract and improve as-is scenarios. Using Simod, SimuBridge allows users to extract as-is process simulation models and resource information from event logs. The logs can be provided by business users. The simulation model discovered by Simod and its parameters are visualized in the user interface (UI). In addition, the UI shows scenario-related parameters (e.g., number of to-be simulated instances), process model-related parameters (e.g., activity duration and their distributions), and resource-related parameters (e.g., roles and timetables). Thus, business experts have an initial simulation model with parameters available and can check their plausibility, adapt parameters where needed, and add the missing information. SimuBridge also allows manually creating simulation models and resource information.
2) Create new, manage, and compare scenarios. Usually, business experts want specific simulation parameters to be changed while keeping the rest of the parameters as in the as-is model. To support this, SimuBridge allows the creation of new scenarios based on existing ones. Additionally, scenarios can be configured from scratch without the need for process mining. Scenarios can also be deleted. Lastly, to see the diferences between scenarios, two scenarios can also be compared to each other, and the diferences between them are highlighted.
3) Start simulation runs. Finally, users can simulate the scenario without having to manually translate it into the specific input formats of a process simulator. Thus, the tool transforms the selected scenario of a user into files needed for the open-source simulator Scylla and provides the simulation output files for download.
BPS 1 0..* Scenario 1 Project 1
1
Resource Parameter Set 1 0..* 1 1
0..* 0..* Resource 0..*
Role 00....** 0..* Timetable 10..1 0..* Process
1 1 1
Model 1 0..* Parameter Set 1 0..* Gateway 0..*
Event Activity
SimuBridge is a web application. As illustrated in Figure 3, its UI comprises various interconnected pages that interact with an internal data storage system, which is based on the browser-side IndexedDB API1 for client-side storage of large structured data.
SimuBridge internally stores the scenario data in a purpose-built data format. Figure 2
provides a high-level overview of this format. It was designed by analyzing the data formats of
diferent research process simulators like BIMP and Scylla and process mining tools like Simod,
and based on an analysis and overview provided in [
To organize several process simulation scenarios, they are grouped into a BPS (Business Process Simulation) project. Each scenario includes a resource parameter set and zero or multiple process simulation models. Some process simulators, such as Scylla, can simultaneously simulate multiple business processes. The resource parameter set contains information about the resources responsible for carrying out the activities specified in the process models. These 1See https://developer.mozilla.org/en-US/docs/Web/API/IndexedDB_API [Accessed: 15/06/23] XES + Simod Configuration
BPMN XML + Simod-specific Simulation
Parameters
Scenario Mgmt. View Scenario
View Model-based Parameters
View
Resource/ Timetable View Simulator
View Internal storage BPMN XML + Scylla-specific
Simulation parameters XML + Scylla logs
resources can be assigned to roles that serve as a mechanism to group resources with similar capabilities and responsibilities. Each role is assigned a default timetable to determine when resources for this role are available. Additionally, resources can have their individual timetables.
A process simulation model consists of a BPMN process diagram in the form of an XML file and the model parameter set. This set describes the execution dynamics of individual process model elements, including activities, events, and gateways. For each activity in a process model, the following details need to be provided: execution duration along with its distribution, activity costs, and the required role(s). An inter-arrival time distribution determines the occurrence of catching events. In the case of XOR and OR gateways, the execution probabilities for the outgoing sequence flows must be defined.
SimuBridge’s UI is structured as follows. The discovery view interacts with Simod to start process mining runs. Along with the user-given event log, in the form of an XES file, it passes a pre-defined Simod configuration. This configuration specifies aspects such as that Simod should discover resource roles, which we can then import to SimuBridge.
Simod [
The scenario management view shows all existing scenarios, allows creating new ones based on existing ones, and comparing and deleting them. The scenario view provides an overview of all parameters of an individual scenario and allows general editing information of that scenario, such as its name.
Additionally, in the model-based parameters view, the BPMN diagram is shown. Users can review and parameterize the individual process model elements, such as activities and gateway. The resource/timetable view allows reviewing the existing resources, their assignment to roles, and the timetables of the availability of roles and resources. Finally, the simulator view allows starting a simulation run for a scenario in the open-source business process simulator Scylla. For this, the simulator view transforms the internal structure into the Scylla-specific files for the simulation parameters. As soon as the simulation has terminated, the output of Scylla can be downloaded.
We evaluated our tool with three diferent publicly available event logs: The Disco tutorial
purchasing example2, the ProM tutorial repair example3, and the BPI challenge 2012 dataset [
To validate our idea from a business perspective, we presented SimuBridge to two business experts with strong business process management and modeling expertise. They confirmed the value of comparing multiple scenarios and emphasized the usefulness of scenario management. They also agreed that a tool with automatic simulation scenario creation greatly simplifies the simulation process because it is dificult to enter hundreds of parameters manually. While acknowledging data imperfections, they still stressed the importance of being able to edit the data manually. Additionally, these experts gave us recommendations on how to further improve the tool, including to support diferent currencies within a single simulation scenario and to make the simulator output directly presentable to the company’s decision-makers.
In the future, SimuBridge can be extended in several directions. First, we want to provide the business experts with more guidance in reviewing their discovered simulation model and improving it where necessary. Furthermore, we want to allow the flexible integration of diferent discovery techniques and process simulators. For this, we must allow the visual specification of which discovery outputs are used for which internal simulation parameters, a form of visual data mapping. With regard to the process simulators, it must be possible to define how the internal data representation is transformed into the input format of the simulator. Specific simulator information might be missing, so the user should be able to add them manually.
Acknowledgments. We thank Bachelor and Master students Tutku Alpsar, Alexander Gottwald, Furat Hamdan, and Andre Schleypen of the TU Berlin for their contributions to the initial implementation prototype.