Process Optimization

[](https://github.com/ramp-eu/Process_Optimization

Contents

• Background
• Usage
• Feedback
• License

Background

Process Optimisation performs process quality and efficiency optimisation using nonlinear model predictive control. The system learns the dynamics of the production process, an then it can predict the process output quality metrics given the current values for the control parameters. The user can also make simulated predictions about the key quality indicators with alternative control parameter values, so as to seek for better control. The system can also find optimal control values given the desired output quality and optimisation constraints, and can thus be used as an advisory tool for the operator or as an autonomous closed-loop controller.

Technical Architecture

ProOpt has three main functional modules and communicates with end-use applications through a RESTful API.

1. The model training module receives training dataset and model specification from external applications to run a training pipeline and results in predictive models. The models are stored in the RAMP storage for later use by other modules.

2. The Prediction & simulation module reads process data from external applications and returns the predictions made by the trained models. External applications can change process data and send the data to this module to simulate the target quality.

3. This Control optimization module is responsible for finding the optimal process setups to achieve the targeted quality. This module receives process data, the optimization constraints and the target quality from the external application and runs the optimization algorithm. The module returns the optimal values of the process parameters, together with the achieved quality

Usage

0. Contact TDS and RAMP for granting access to private docker repository in RAMP (docker.ramp.eu)
1. Clone or copy folder docker to your local environment
2. Login to docker.ramp.eu through docker CLI
3. Modify `docker-compose.yml' to mount local folder for AI model storage after training
4. Exec docker-compose up -d

The API docs will be then accessible via http://localhost:6543/api/v1.0/.

Notes:

• For tunning the training and optimization, please refer to documentations under docs folder.
• For making requests to the API, please refer to this example in python, docker/examples/main.py.

Definition of the API interface:

Information about the API can be found in the [API documentation](http://localhost:6543/api/v1.0/) of the running docker container; or `src/openapi.yaml`

Feedback

Any feedback and suggestions can be submitted by creating New issue in the Issues tab or by emailing the development team:

• Lauri Mannila ([email protected])
• Juho Piironen ([email protected])
• Duc Ta ([email protected])

License

Proprietary ©