StarCS is an open, plant-wide control platform with a wide-ranging and versatile set of tools aimed at process stabilisation and optimisation. Using the latest technology in both software and advanced process control, Mintek has developed an adaptable and operator-friendly control system.
Although StarCS was developed specifically for the metallurgical process industry, it can be used to control any industrial process.
StarCS incorporates a layered approach, which enables implementation of just about any controller. The versatility of this control platform stems from the use of OPC communication, an International standard procedure for the process industry.
ROBUST NONLINEAR MODEL PREDICTIVE CONTROL
The StarCS Robust Nonlinear Model Predictive Controller (RNMPC) was specifically developed for problems associated with the minerals processing industry. Its particular characteristics enable the handling of integrators, noise, model errors, limits, interaction, widely differing dynamics, and an unequal number of manipulated and controlled variables.
The MPC uses an objective function and multivariable control technique to calculate the future movement of all available manipulated variables to eradicate future deviations from setpoints in the controlled variables. The MPC is very capable in handling multi-input-multi-output (MIMO systems), since it uses an optimisation function that does not need equal numbers of manipulated and controlled variables to calculate the best possible movement of the manipulated variables. This feature also makes the MPC very efficient in handling sensor or equipment failures.
REAL TIME OPTIMISATION
Having a mathematical model of a process presents endless opportunities – from implementation of a model based controller and optimiser, to simulating the process for the purpose of training operators. However, obtaining such models are challenging and time consuming. Mintek has therefore developed the automodeller which perturbs a process unattended and uses the frequency response data to fit process models.
The StarCS Autotuner is a module that was developed to ensure continual optimal plant performance. It provides the plant with a tool to tune their controllers whenever necessary. The Autotuner automatically adjusts controller parameters of the control loops on a plant online. This ensures enhanced plant performance in the short and long term.
The Autotuner is also used by the Mintek engineers to efficiently commission our control systems, shortening the time required for the commissioning.
Many metallurgical plants are hampered by slow communication networks and long deadtimes, with communication loop times of up to 5 seconds, but at the same time having fast process and filter dynamics. PID Control is generally ineffective in controlling such plants.
The StarCS Deadbeat Controller implements an algorithm that overcomes this problem and at the same time is robust to plant noise and disturbances as well as model uncertainties. It is based on advanced discrete controller design techniques.
Fuzzy Logic provides a framework allowing the effortless incorporation of process know-how into a comprehensive control strategy. Its main strengths are shown in the absence of reliable measurements or models, and where the art of controlling a process is defined better using ill-described concepts rather that crisp rules. Rules and functions are set up in a logical and intuitive fashion, allowing rapid controller development and integration.
The Simulators module has an extensive suite of simulators that can be used to easily simulate almost any plant. Linear models of any order can be simulated for both single and multiple inputs/output simulators.
Other features such as noise, limits, drift rates, actuators with slack, local control loops on inputs and additional transfer functions on outputs are also supported.
This tool is essential for operator training and understanding how to interact with other Mintek measurement and control products.
COMING UP . . . DEEP NEURAL NETWORKS
We are currently working on developing a deep neural network that is able to perform control, optimisation and tuning of systems.
Watch this space for more…