The probability of default on loan repayments, the risk of rain, and even climate change scenarios can nowadays be predicted very accurately using simulations. The basis for this – in addition to powerful computers – are models. They are mathematical representations of reality. From the mathematical laws, forecasts of future behavior can be derived. As early as 1814, the quest for the most accurate possible simulation of the physical world culminated in a system of mathematical world equations established by the mathematician Pierre-Simon Laplace.
Laplace put forward the following thesis: A superior mind, which knows the state of motion of every atom in the entire universe with knowledge of all the laws of nature, must be able to predict the future quantitatively. This superior entity, later referred to as the ‘Laplace demon’, is based on a strictly deterministic world view. It was later caught up by the theory of relativity, Heisenberg’s uncertainty principle and quantum theory. And yet: the prediction of future events is as exciting today as it was 200 years ago. But it doesn’t have to be the entire world whose future is being predicted – extracts of special events would be spectacular enough, wouldn’t they?
Stationary operations
Predictions have always been an important part of scientific methodology. Those who can predict how reactants behave depending on the process environment understand the process and thus master it. Process engineering focuses on mathematical models of physical, chemical and thermal processes.
Accurate modeling provides the process industry with the key to optimized processes.
Thus, simulations for stationary
processes have been part of the basic tool box of process engineers for many
years. However, such methods only cover a small part of real operation and
their ability to make predictions is limited. If it could be determined with
sufficiently high probability when
- the filters in reverse osmosis plants should be
replaced,
- the so-called fouling in heat exchangers makes
cleaning necessary or
- coking in steam crackers exceeds a critical value,
the processes concerned could be
made far more efficient, resource-saving and environmentally friendly. This is
precisely where equation-oriented process modelling functions come into play.
A more sophisticated process development environment
The modelling of chemical processes
is usually done in two ways:
1. using
sequential oriented modeling (SM)
2. with
the equation oriented approach (EO)
With SM, the simulation is carried out in the order of the defined inlet flows and is therefore quite robust and easy to handle. With the EO approach, on the other hand, a system of equations is created over the entire flow diagram and solved at once. EO is also particularly suitable for optimizations and model refinements. Modeling environments such as gPROMS from PSE (Process Systems Enterprise, now a Siemens company) have been specially developed for this purpose. With their help, even complex models, as required for dynamic process simulation, as well as for off- and online real-time optimization, can be solved. Until now, the EO approach has been numerically more challenging and is therefore not yet widely used as a standard. PSE’s pioneering work over many years has successfully answered these and other challenges with Model Initialization Procedures (MIPs). Thanks to MIPs, models can be solved automatically without preset initialization parameters. This combines the advantages of SM and EO into a robust and easy to use tool: gPROMS ProcessBuilder and gPROMS FormulatedProducts for advanced process modeling.
Simulation models for all life cycle phases
With such highly developed
simulation models, operators benefit in every phase of the life cycle of their
plant – from planning simulation, simulation-based engineering with virtual
commissioning and operator training to real-time process optimization during
operation. The planning simulations from the conceptual design can be reused as
a basis for all subsequent simulations. Today, it no longer uses only
stationary but also dynamic models to design the process, including start-up
and shut-down behavior, more efficiently.
Process engineering is followed by automation engineering. Here the modelled plant behavior can be used to test the hard- and software engineering. In addition to the process engineering, the device behavior must also be modelled in detail here. As a rule, simplified, but real-time capable dynamic process models are needed in this phase. This can be easily realized with Simit. The simulation platform from Siemens represents the field devices that communicate with automation level. The automation program created can then either run on real controller hardware and be tested against simulated field devices (hardware-in-the-loop) within the framework of virtual commissioning. Or a computer emulates the automation (software-in-the-loop). In both cases, all automation functions can be safely tested before actual commissioning. In this way, design errors are detected and corrected much earlier and the automation programs can later go into actual operation with a near-zero error rate. Parallel to virtual commissioning, simulations also serve to prepare plant operators for their jobs: In operator training environments, personnel can be trained with Simit. The existing high-precision dynamic process model, for example from gPROMS, can be coupled with Simit (so-called co-simulation).
Forecasting and operational optimization
Until commissioning, the process simulation is enriched with process engineering knowledge, and from the start of operation the model from the early process design can also be coupled with real data. In this way many things can be realized today that were not possible a few years ago. So-called soft sensors, for example, calculate non-measurable process variables based on the model and taking into account existing measured values. But the simulation model does much more: Since it is not bound to physical or time limits, it can be used in combination with real measured values – both real-time and historical data – for predictions. These can be, for example, runtime predictions: at which date filters need to be changed or maintenance has to be carried out to keep up optimal performance.
However, it is also possible to simulate operating points with regard to different modes of operation. Operators thus receive very precise answers to questions such as: “How must my plant be operated in order
(a) to
run maintenance-free for as long as possible?”
(b) to
obtain as much end product as possible in a short time?”
(c) to
drive as energy-efficiently as possible?”
Such and countless simulated
scenarios lead to operational optimizations in real-time and are of inestimable
value. As you can see, it is not necessary to keep a demon in the plant to take
huge steps forward!
Survey: What’s your opinion?
What’s your opinion on the use of simulation? As part of a joint research project from Siemens AG, Karlsruhe Institute of Technology (KIT), Technische Universität Dresden, Hochschule Pforzheim and the VDE/VDI GMA Fachausschuss 6.11. a global survey has been developed. Just spend 15 minutes and help shape the future of simulation! The survey is completely anonymous and will be available until July 15th. For taking part in the survey just click here: https://bit.ly/survsim.
I’m looking forward to your participation!
