A simulation framework, on which models, processes and controllers can be easily, flexibly and concurrently simulated. Includes automatic real-time visualization of process variables and an interactive CLI. Intended for educational use, for people interested in learning about modelling, simulation, control and optimization.

These instructions will get you a copy of the project up and running on your local machine for development and testing purposes.

This example will run a Continuous Stirred Tank Reactor (CSTR) model. A MPC Controller will aim to steer the manipulated variables in order to reach the control objectives (maximizing product yield). The example model is available here, the controller is defined here.

#Clone the project
git clone https://github.com/marianaipmonteiro/ProcessControllerSimulator

#Change into project directory
cd ProcessControllerSimulator

#Create a virtual environment
python3 -m venv venv

#Activate the virtual environment
source ./venv/bin/activate

#Install Dependencies
pip install -r requirements.txt

#Run Main, providing a simulation class as an argument
python3 Main.py examples.simulations.cstr_control_simulation.CSTRControlSimulation

Connecting to localhost:8050 should present the automatic real-time visualization of process variables:

A CLI is provided which allows several interactions with the running simulation:

# Help message
Commands:
            print world - Prints the state of the world.
            perturb <var> <value> - Perturbs variable <var> by the decimal amount <value>, positive or negative.
            help - Prints this message.
            exit - Exits the program.

The main abstractions in the framework are that of a WorldState, which contains the current state of the world (constants and variables). A Simulation owns a list of WorldStates, and will use a WorldInitializer to create the initial world, providing definitions for constants and variables. A Simulation is initialized by providing it with a number of SimulatedSystems.

Each SimulatedSystem is instantiated by the Simulation, and runs a simulation loop on a separate thread. This loop will periodically poll the WorldState list, obtaining the most recent WorldState, which is used by the SimulatedSystem to compute a new WorldState, and appending it to the WorldState list, thus progressing the state of the world. A simple lock is used to provide exclusive acess during accesses to the world state list, but users are abstracted from that.

Both Models and Controllers are SimulatedSystems, with specific simulation loops. To create a model, users may extend either Model.py or SelfInitializingModel.py (which is both a Model and a WorldInitializer). An example Controller is provided in the form of an MPC Controller.

Visualization is achieved by periodically rendering the values of a variable in all the WorldStates over time.

• Flexible Simulation Framework
• Initial Examples
• Visualization
• Logging
• Figure out why Data Visualizer takes so long
• GUI for Simulation Customization
• Reinforcement Learning Controller Framework
• Surrogate Model Framework
• Benchmarks
• Documentation
• Testing

Mariana Monteiro - @marianaipmonteiro - Modelling the CSTR, implementing MPC strategy, computational model.

Pedro Silvestre - @PSilvestre - Computational model.