BASF SE – Predictive Decision Support for Increased Efficiency in Batch Plant Operations

Goals

• Harness INOSIM digital twins to maximize situational awareness of plant operators
• Provide predictive decision support at unprecedented levels of quality
• Maximize the efficiency of operator movements and actions within the plant

BASF creates chemistry for a sustainable future and combines economic success with environmental protection and social responsibility. More than 110,000 employees in the BASF Group contribute to the success of its customers in nearly all sectors and almost every country in the world. Digitalization is a key element of BASF’s new corporate strategy and presents big opportunities also in the field of Global Engineering at BASF on its numerous production sites. This success story describes a project which has been a part of a larger lighthouse digitalization effort that BASF SE is executing for one of their batch production plants at their integrated production site in Ludwigshafen, Germany.

Industrial batch production represents an important part of the production capability of BASF SE. Batch plants are complex systems in which dozens of products may be produced in parallel, each using possibly dynamic production recipes with many production steps. In addition, batch production takes place in a dynamic environment (e.g. varying market conditions, strict time-to-market requirements, or unforeseen disturbances and downtimes). This level of complexity can make these plants somewhat “opaque” to the operations personnel and production managers, as it can be nearly impossible to predict the future state of the plants over long periods. Thus, decisions must often be based on intuition instead of facts.

Being aware that “knowledge is power”, the main goal of the team of Global Engineering at BASF SE was to employ a highly accurate simulation model implemented in INOSIM’s Simulation Software to predict the behavior of batch production plants much more accurately than it was possible before, as such accurate predictions create transparency and situational awareness that allows the plant operators and management to operate and plan much more efficiently.

Decision support that is based on accurate real-time predictions provides a variety of potential benefits: Resource and personnel planning can be made much more accurate and less conservative than before; the effects of delays, errors, disturbances, or deviations from the best possible production can be counteracted before they occur; important KPIs can be predicted days or weeks in advance; maintenance can be planned based on accurate predictions of suitable maintenance slots. This case study focused on accurately predicting manual interactions throughout the entire production plant, such as sampling, filling, change of catalysator, etc.

Challenges

• Build a highly accurate digital twin of the real batch plant
• Connect and align the INOSIM digital twin with the real batch plant
• Build a robust production-near execution environment for predictive simulation
• Successfully deploy the system into production and ensure acceptance by the plant personnel

To be able to use an INOSIM simulation model for predictive decision support, the Global Engineering team at BASF SE had to solve several challenges.

They had to build a simulation model of the batch plant that accurately predicts those dynamics and behaviors that are needed to answer the questions that operators and managers are interested in. Since all predictions must be made starting from the complete state of the real batch plant, they had to modify the simulation model such that it can be initialized to this state of the real batch plant at any time.

They had to develop the capability to connect the simulation system to their IT/OT infrastructure to enable them to import the measurements and data signals in a format that can directly be used for model initialization.

They had to add algorithms to the simulation model that extract and pre-process the relevant information from the plethora of simulated results data that is generated by complex simulation models, and they had to find ways to transfer this information into visual dashboards that can be accessed by plant personnel.

To tie it all together, they had to develop the capability to parametrize and robustly execute the complete toolchain (data retrieval from the IT/OT side, data transfer, model initialization, simulation, information extraction, and visualization) unattended in a production-near IT environment.

Solutions

• Built a highly accurate digital twin that can be driven to the state of the real batch plant
• Developed a software architecture around the INOSIM simulation software that realizes the complete toolchain (data retrieval, simulation, visualization) for predictive decision support
• Deployed the system into production and connecting it to OSIsoft PI (for data retrieval) and Power BI (for visualization)

First, the team of Global Engineering at BASF SE built a detailed simulation model of the plant using the simulation software INOSIM Expert Edition. The plant consists of three sub-plants, each of which realizes a complex production process involving several reactors and feedback streams. The model was validated at the real batch plant.

In a second step, BASF SE developed customized algorithms that allow them to drive the simulation model into any state that the real plant can also assume. The algorithms were implemented in INOSIM’s powerful VBA-based scripting engine which can add custom algorithms to any part of an INOSIM model, thereby providing virtually unlimited capabilities for model customization.

In cooperation with the team of BASF SE, INOSIM developed a software architecture around the INOSIM simulation software that enables BASF to realize the complete toolchain for predictive decision support. The architecture was designed for unattended robust execution of complex simulation toolchains over long periods, with extensive logging and error handling capabilities to ensure that any irregularities during execution can quickly be identified and mitigated.

The Workflow Controller component was implemented as a Windows service program that regularly executes the simulation toolchain reliably in production-near IT/OT environments. Most of its execution parameters can be customized, such as frequencies or the structure of the toolchains. For the import and pre-processing of plant data sets, a scalable data integrator component was developed that provides an open plug-and-play platform for data connectors that can access a variety of different sources. The first connector was developed in this project to connect to the PI system by INOSIM partner company, OSIsoft, that BASF SE employs as a data management system for the batch plants of this project. The data sets to be imported and their mapping to data entities within the simulation model are specified in easy-to-understand configuration text files that the user can write in any text editor. The retrieved data sets are transferred to user-specified pre-processing functions that prepare the data for ingestion into the simulation model. After ingestion, the data sets are seamlessly accessible within the simulation model via the integrated scripting engine and were directly fed to the custom initialization routines developed by BASF SE.

How INOSIM and BASF SE built an environment for predictive decision support around the INOSIM simulation software suite.

Information extraction from the simulation result data sets as well as pre-processing and transfer to the employed business intelligence solution Power BI for visualization was also implemented within the native INOSIM scripting engine. An IT infrastructure was created by the team of BASF SE, enabling the required access to the underlying systems to operate a real-time simulation solution. Plant and lab operators can visualize the results of the prediction on an intuitive dashboard either direct on a screen in the control room or from any mobile device.

The software architecture that INOSIM developed in cooperation with BASF SE is the basis for the Foresight System For Predictive Decision Support that INOSIM is currently developing and quickly driving towards commercialization. In addition, INOSIM BICON (Business Intelligence connector), an INOSIM add-on that provides a seamless, highly customizable transfer of simulated and custom data sets from INOSIM simulations to the business intelligence solution Tableau, was inspired by this project. It will be released commercially within the next months by INOSIM.

Screenshot of a productive dashboard that provides an overview of the upcoming manual tasks and interactions for the next hours. In addition, the dashboard provides information about the current delays of the reactors.

Results

• Successfully established the new system as an essential tool for the planning and prioritization of manual interactions and tasks by the plant personnel
• Achieved full return of investment within four months
• Established INOSIM as a highly suitable tool for predictive production support
• BASF SE aims to deploy the system to other plants

The new predictive decision support system was deployed into operation in July 2020. With its support, operators and lab personnel are now much more efficient in the planning and prioritization of manual interactions and tasks than before. As an added value, the system provides information about delays of the reactors, significantly improving the understanding of the real plant state for the operators.

The predictive decision support system has been well accepted and considered to be very supportive by the operators of the plant and by its lab personnel. A return on investment of 100% was expected after four months of operation. The projected business case for the realization of this project has been confirmed by the plant director after approximately six months of operation of the installed solution. The successful realization of this project has confirmed INOSIM simulation models as a highly suitable tool to make accurate quantitative predictions for batch plants, opening a wide range of valuable decision support potential that BASF SE currently proofs for further plants.