Process Modeling and Digital Twin: Significant Benefits — a Whole Life Long
The digital twin of a plant is a virtual image, which not only allows internal process states to be displayed transparently, but also allows process optimizations to be worked out. Take the example of steam crackers — the king’s class of process plants — and learn how models, simulations and digital twins can be used to generate benefits over the entire life cycle of plants.
Model-based technologies are the key to simulation and most optimization measures to model real-life scenarios by applying mathematical equations. A digital twin is an image that reflects the real system with maximum possible precision — including all the system components, as well as its features and functions.
When regarding the digital twin of a process plant as a consistent concept, it turns out that there are actually several digital twins: the digital twin of the product, the digital twin of the production plant and the digital twin for modeling product and production performance. The following principle applies to all models and simulations and is thus essential to understand the digital twin concept: the appropriate level of detail (i.e. the modeling depth) depends on the individual purpose and context. Cost-benefit considerations play a major role in this — the more precisely a model reflects real-life conditions, the more complex and cost-intensive it will be. This means that a digital twin’s scope of functions is strongly determined by its intended purpose.
This scope can vary significantly from process plant to process plant — ranging from production simulation to production process optimization up to economic aspects. Simulation modules are already being used in all the individual disciplines today. However, the following approach now opens up entirely new perspectives: combining the individual models and software tools into a consistent, semantically coupled system — across all of a plant’s many hierarchical levels and life cycle phases. What is so promising about this approach is the fact that distributed and central information bases can be coupled via so-called exchange and co-simulation standards.
The following example demonstrates how models, simulations and digital twins can nowadays generate significant benefits across a plant’s entire life cycle.
Steam Crackers — the King’s Class of Process Plants
Steam crackers are among the most complex plants in the petrochemical industry. Due to the high throughput rates and the great economic significance of a cracker, plant operators want to make sure that they leverage the full optimization potential and minimize the number of unavoidable plant standstills, e.g. for cleaning work (de-coking). To achieve this, the following two special challenges must be overcome:
- Considering that the yield of primary products depends on various influencing parameters, a suitable multi-variable control procedure is the best approach. However, the yield cannot be measured directly on the cracker furnace outlet, but only for several furnaces altogether and with a considerable delay after the cool down. This means that the target variables should preferably be measured at a point in the process where, however, no measuring device could survive in the long run. A soft sensor is extremely useful for this because it allows to replace a real sensor with software based on a process model that allows real-time simulation parallel to operation.
- As a result of high temperatures, the slow deposition of coke inside the tubes is unavoidable (referred to as coking). This degrades the efficiency of the steam cracker unit. The actual level of coking should be known in order to be able to schedule cleaning (de-coking). Visual inspection during operation is not possible. To predict the residual service life, future scenarios can be simulated based on the current state, thus becoming available faster than in real time.
To model a cracker with the precision required for such predictions, a total of more than 10,000 coupled differential and algebraic equations must be applied.
Simulation Right from the Early Stages
A great deal of in-depth process knowledge is required to create such models. Starting with the first conceptual considerations concerning the plant design, this knowledge is nowadays aggregated across numerous software systems. Based on existing plant knowledge and recent findings presented in specialist publications, a first digital twin of the process can thus be created through simulation software. Afterwards, this digital twin serves to determine the plant layout and the components required (so-called “conceptual design”).
For a cracker, for example, this includes determining the chemical reaction, as well as the optimal reactor sizes and wall thicknesses. All the knowledge obtained during this phase is combined in a Process Flow Diagram (PFD) — the basis of the digital twin of the process. The equation-oriented process modeling during this stage is performed by the gPROMS ProcessBuilder — a modeling and solution environment achieving higher modeling precision than any other flowsheet simulator.
This sophisticated technology was developed by Process Systems Enterprise (PSE, London) and will be integrated into the Siemens digitalization portfolio after the acquisition of PSE. The detailed process simulation created by gPROMS can be used across the entire life cycle.
During the further course of engineering, the digital twin of the process will be transferred into the plant planning tool: It thus provides the basis for the digital twin of the plant which will be extended step by step by further plant-relevant aspects such as sensors, actuators and control structures. The Comos plant planning tool from Siemens is ideal for creating the basic structure of the digital twin of the plant in a single, object-oriented data model — from the process flow diagram up to the complete R&I scheme. The digital twin of the plant in the plant planning tool is extended by the automation components — such as the process control system, process instrumentation and operator stations — and the structural planning is enriched by detailed engineering information. Once the process engineering has been completed, all the necessary information is transferred to the engineering system of the distributed control system — ideally a Simatic PCS 7 from Siemens. Furthermore, the field level is mapped on the Simit simulation platform as a prerequisite for the virtual commissioning of the automation software.
From Planning to Operation
The goal of virtual commissioning is to obtain an automation system that has been tested as completely as possible. The focus is on testing the control programs implemented — which are unique for each plant. A simulation model can be used to test the real control programs. Addressing the entire communication interface between the automation components and the field, this model is connected to the real (Hardware-in-the-Loop) or emulated (Software-in-the-Loop) control hardware. Virtual commissioning with Simit allows testing all the automation functions in a safe environment prior to real-life commissioning: All the systems, machines and processes are simulated on the basis of existing planning and engineering data as well as specific libraries. The fact that it is no longer necessary to create every single model is indeed a major benefit: The model evolves over the entire life cycle in order to include various sub-models.
Parallel to checking the automation technology, this simulation technology is also ideal for preparing the plant operators for their future job: Simit can be used as a virtual training environment (Operator Training System, OTS) in order to train the plant operating team before commissioning — on real operator displays and with original automation programs. To minimize response times, the plant personnel can thus be trained for both normal operation and fault scenarios. In the case of the OTS as well, the level of detail plays a major role. A “high-fidelity OTS“ e.g. is required to obtain a highly detailed process model. Instead of creating it from scratch as was the case so far, a so-called co-simulation, with coupling of gPROMS to Simit, is now possible in the context of the digital twin.
Optimized Operation and Maintenance
Let’s get back to the two major challenges of operating a steam cracker: virtual target variable measurement via soft sensor and coking. Soft sensors represent an important application field of the digital twin during the operational phase: To estimate a missing process value, a soft sensor relies on a process model and other available measured variables. The digital twin of the plant should be used in order to avoid having to create a new model for each application of a model-based soft sensor. However, implementing a soft sensor is worth the additional effort if the estimated value provided is relevant for process control. The yield of steam crackers can only be measured several steps downstream in the process. Due to the dead time — which exceeds the actual process dynamics by far —it is impossible to directly control the measured yield. However, the soft sensor can provide an estimate of the yield at the cracker furnace outlet (without dead time). This estimated value can serve as a basis for direct feedback control.
The longer the cracker is in operation, the more coke will be deposited. The deposition of coke during production continuously alters the plant’s behavior until the coking reaches a point where the cracking furnace must be shut down and de-coked. To minimize downtimes for de-coking, the operating parameters should be adapted such that the intervals between de-coking are as long as possible. The investment in the digital twin of the process during the plant design phase now pays off once again: The simulation model is simply ideal for carrying out “what-if” experiments.
These scenarios are extremely useful for the plant operator to e.g. maximize production times (residual service life until de-coking) and optimize maintenance scheduling. In the case that production conditions change due to a resource bottleneck or volatile raw material prices, such experiments with a digital twin of the process can help to improve the economic balance.
Summary: Model-based technologies are the key to simulation, process optimization and even secure predictions. The consistent use of a digital twin maximizes the economic benefit across a process plant’s entire life cycle — and even more so because new simulation models do not need to be created for every step, but can be coupled or merged.