Model-based Control is suited to optimization of well-understood unit processes (SAG Mills, Ball Mill Circuits, Flotation Circuits, Thickeners as an example). The performance is superior to that of Model-free systems (Expert Systems as an example) because they are capable of anticipation and thereby can predict the process response to new situations.
Optimise your data assets. Make faster and better decisions. Discover insights into your processing plant. Find new sources of opportunity for throughput or yield improvement. Capitalise on the untapped intelligence you already own. INNOVATION [X] offers bespoke solutions, consultation and Software as a Service (SaaS) solutions such as the following.
DYNMICA monitors the constraints throughout the processing plant in real-time. This capability identifies the total plant utilisation allowing for significant improvements in process plant throughput.
OCTOGRAPH (O.) is a control performance assessment and design optimisation platform. O. has been applied to industrial applications globally to assess the current performance of the control schemes installed within operational processes. O. is non invasive to the process, resulting in a plant-wide control performance assessment without lost production or process interruptions.
Point Cloud is a computational engine that transforms image datasets into point cloud representations. The service is low cost on-demand offering removing the need for drone survey companies to own expensive computational software and infrastructure.
Our consulting approach is rigorous and application based. Our approach incorporates four stages, design, simulation, implementation and evaluation. We add significant bottom line cash flow value by implementing advanced control solutions to complex and challenging problems.
The Design Phase of the INNOVATION[X] approach develops control schemes supporting the established objectives for the plant.
The Simulate Phase of the INNOVATION[X] approach validates the developed control schemes by applying a specific control simulation environment to the design. This phase allows for disturbance and control performance testing.
The Implement Phase of the INNOVATION[X] approach interprets the validated control schemes to logic specific to the client’s control system environment. Commissioning and troubleshooting activities are also incorporated within this stage.
The Evaluate Phase of the INNOVATION[X] approach evaluates the implemented control schemes against the business drivers to confirm addition to bottom line cash flow value.
Our focus areas of technical expertise and implementation services are:
Operational facilities, whether it manufacturing, mining, or petrochemical, process variability is potentially detrimental to the profitability of the business. Advanced Control techniques can allow reduction in process variability and provide the ability to PUSH plant constraints.
Continuous time simulations are developed to functionally test control schemes by injecting disturbances of significance to fundamentally test the rejection ability of the proposed scheme.
To understand and map the relationships between elements, enables the ability to construct visualisation tools to see process characteristics from a different behavioural view.
To simulate; is to answer questions. Questions of a process, performance limits, asset dynamics or flowsheet characteristics. A simulation will build knowledge.
What is the “best” way to operate this plant? What is the “best” way to co-ordinate our fleet? What is the “best” solution to the question? By “best” we mean OPTIMAL.
The ability for a MACHINE or SYSTEM to learn from a data set has unlimited opportunities and applications. We are unlocking this potential with real-time industrial applications.
Estimation (or estimating) is the process of finding an estimate, or approximation, which is a value that is usable for some purpose even if input data may be incomplete, uncertain, or unstable..
A conceptual model is a representation of a system, made of the composition of concepts which are used to help people know, understand, or simulate a subject the model represents.
An understanding of the control performance of a process plant is the first step in realising production or performance increases.
The application of process unit and circuit / plant - wide control is challenging. Our experience across varied machine, process units and mineral processing circuits is extensive. Examples of our breadth of experience are:
The objective of cyclone pressure control is to have a consistent pressure at or near the manufacturer’s recommendations for classification efficiency for a variety of feed flows and densities. A relatively constant cyclone feed pressure is maintained by adjustment of the number of cyclones turned on in the cluster. The cluster is assumed to have of the order of 10 to 15 cyclones. Because of the quantisation effect of selecting a whole number of cyclones in service, it is impossible to keep the pressure completely constant. However, variation in pressure does not greatly affect cyclone d50c, so the small variations in pressure that remain have no significant affect on cut size or ball mill circuit control.
Sound is assumed to be a measure of the breakage conditions at the impact zone of balls (and to a lesser extent, rock) in the mill. High sound levels occur when the ball strike point distribution is high on the liner wall and not overlapping the mill charge toe. Low sound levels occur when the strike point distribution dominantly falls within the toe of the charge (and is damped).
The microphones for the sound level are typically placed as close as possible to the mill shell at the likely strike point of the balls with a fully charged mill. This closeness helps reduce interference by external noise sources such as crane “hooters” etc. In order to emphasise the ball strike component of the sound level, the microphone electronics will often apply a band-pass filter with a centre frequency of approximately 1 kHz to the signal. There are also more advanced sound instrumentation options that can also produce a signal that needs to be kept at a setpoint to maintain optimal breakage conditions for any set of mill states (e.g. load). Holding the sound level constant is, by inference, holding the breakage conditions constant at the desired setpoint. Depending on mill design, the setpoint may be a compromise between best breakage conditions (high sound level) and limiting liner damage due to ball strikes (low sound levels). For any particular mill filling, there is a mill speed which will maintain the setpoint sound level (and therefore grinding conditions).
The process objective of the ball mill circuit control scheme is to maximise the grinding effort applied to solids delivered by the SAG mill. To prevent spillage and overload, the control scheme must allow the ball mill circuit to accept almost any amount of feed rate without overload. At high feed rates, this will compromise the grind size, so the SAG mill feed rate limits must be managed to ensure the right grind size versus throughput tradeoff is achieved. This can be done automatically if a cyclone OF particle size analyser is available. At very low feed rates the ball mill circuit should not overgrind the ore, despite the grinding capacity being available to do so.
Dynamic control of the grinding circuits do not involve any substantially multivariable characteristics with strong interactions. In order to maximise throughput or grinding performance, the equipment needs to run at its operating constraints. For the subcircuits, only one constraint will be active at a time (1 constraint loop). There are no multivariable interaction effects and the process is dominated by Multiple Input Single Output (MISO) and not Multiple Input Multiple Output (MIMO).
In many grinding circuits, maximum performance (grind or throughput) is achieved at a constraint of the equipment. In this case no formal optimisation is required to set the operating setpoints of the ARC scheme. Only constraint control is required (which is contained within the ARC scheme) along with the corresponding equipment limit setpoints.
The objective of the optimisation component of the grinding circuit control hierarchy is to provide optimal steady-state setpoints for the Advanced Regulatory Control (ARC) layers of the control scheme. That is, the optimiser provides the correct steady-state operating point for certain variables of the plant, and the ARC provides the fast, real-time dynamic control of the equipment to its constraints and to achieve those optimal operating points.
There are a limited set of process variable setpoints which may need optimisation. For maximising the throughput of the grinding circuit, most of the setpoints of the grinding ARC schemes relate to fundamental equipment and process constraints. Maximising the performance of the grinding circuit is done by pushing the equipment up to these constraints. These constraints don’t change unless the equipment or flowsheet changes. However, in general, there are some plant variables which optimal setpoints which need to be adjusted in real-time. The set of these variables needs to be determined after analysis of the ore, flowsheet and control schemes, taking note of whether sufficient degrees-of-freedom (or manipulated variables) are available to achieve these setpoints, after the ARC has “consumed” most of them to maximise real-time performance.
Solid-solid separation processes are used to upgrade a variety of materials such as coal, tin, iron ore, bauxite and heavy mineral sands. Solid-solid separation processes are routinely used to separate minerals based on differences in intrinsic properties such as specific gravity, magnetic susceptibility, or conductivity. The optimisation of solid-solid separation processes based on the concept of constant incremental grade has long been recognized in the technical literature (Mayer, 1950; Dell, 1956). (Luttrell, G.H., Mankosa, M.J., Strategies for the Instrumentation and Control of Solid-Solid Separation Processes, Mineral Processing Plant Design, Practice and Control, Volume 2.)
INNOVATION[X] has developed and implemented model-based control schemes yielding significant throughput improvements. The implementations are robust leading to sustainable long-term production improvements. Demonstrated throughput improvements ranging from 38% to 125% are common.
Flotation process control is a challenging and important task in the ore benefication chain. The efficiency of the flotation process largely controls the economics of the overall mineral processing plant (Hodouin et al. 2000).
Flotation plants are difficult to operate. Non-linear dynamics, coupling among control loops, large and variable dead times, strong and continuous unmeasured input disturbances, imperfect knowledge of the phenomenology of flotation, impede process control. Frequent lack of appropriate and precise instrumentation makes supervision and control even more difficult. (Osorio, Perez-Correa and Cipriano 1999).
A universal way to control a flotation circuit cannot be given. Each circuit has its special features in terms of cell configuration, instrumentation, ore and chemistry, which have led to a large number of different control strategies and methods used and reported in literature. The only undoubtedly common feature between flotation plants is to maximise profit.
INNOVATION[X] has developed and implemented advanced control schemes that manage the cell process conditions. The real-time optimisation offering, Estimata, then maximises flotation circuit recovery.
The general objectives for thickeners and clarifiers are to produce clean overflow and maximum solids concentration in the underflow. Flocculants are typically used to agglomerate the solids to increase the settling rate and improve overflow clarity. Thickener control involves a number of complexities and variables such as varying feed characteristics, changes in feed concentration, solids specific gravity, particle size distribution, pH, temperature and reaction to flocculant which can all contribute to variations in performance.
INNOVATION[X] has developed and demonstrated control schemes that manage the solids settling rate and solids inventory in real-time while managing the known constraints within the equipment and associated processes.
The broad objectives of the Model-based Control Strategy is to maximise the economic return from each cell in the potline by achieving the optimal combination of: 1) Cell power efficiency (maximise) 2) Cell life (maximise) 3) Production rate per cell (maximise) 4) Manual intervention of operators (minimise). INNOVATION[X] has developed a model-based control scheme for a Reduction Cell. The model manages the objectives of the reduction cell while maximising the economic business drivers.
Using a series of rigorous authentic examples, the authors demonstrate several simple yet practical techniques for utilizing adaptive neural networks to produce more efficient process control. Their in-depth description of implementation issues offers a wealth of pragmatic wisdom.
Mills, P.M., Zomaya, A.Y., Moses, O.T. (1996) “Neuro-Adaptive Process Control: A Practical Approach”, Edition 1, Wiley, London.
Mills, P. M., (1988)
Advanced Control of the Alumina Calcination Process
Master of Engineering Thesis, University of Queensland.
Mills, P. M. (1994)
Adaptive Control of Non-linear Processes using Neural Networks
PhD Thesis, University of Western Australia.
Duffy, G., (2002)
Depth Control of an Underwater Platform
Bachelor of Engineering Thesis, University of Southern Queensland.
Duffy, G. W. (2018)
OCTOGRAPH: A Methodology for Returning the Industrial Plant to the Desired Operating Point
PhD Thesis, Griffith University.
McIntosh, P., Greenhalgh, R., and Mills, P. (1987)
“Advanced Control Techniques for Alumina Calcination Rotary Kilns”
A Practical Study of Adaptive Control of an Alumina Calciner
Mills P. M., Tadé M. O. and Zomaya A. Y. (1993)
“Identification and Control using a Hybrid Reinforcement Learning System”
International Journal of Computer Simulation – Special Issue on Intelligent Simulation
for High Autonomy Systems.
Mills P. M. and Zomaya A. Y. (1993)
“A Neural Network Approach to On-line Identification of Non-linear Systems”
Cybernetics and Systems 24, 171-195.
Mills P. M., Zomaya A. Y. and Tadé M. O. (1993)
“Adaptive Model-Based Control using Neural Networks”
International Journal of Control.
Mills P. M. and Zomaya A. Y. (1991)
“A Neural Network Approach for On-line Identification of Non-linear Systems”
Proceedings of the International Joint Conference on Neural Networks 1, Singapore,
Mills P. M. and Zomaya A. Y. (1991)
“Reinforcement Learning using Backpropagation as a Building Block” Proceedings of the International Joint Conference on Neural Networks 2, Singapore,
Mills P. M., Tadé M. O. and Zomaya A. Y. (1992)
“A Hybrid Reinforcement Learning System for Identification and Control”
Proceedings of the AI Simulation and Planning in High Autonomy Systems
Conference, Perth, 2-8.
Mills P. M., Tadé M. O. and Zomaya A. Y. (1992)
“Adaptive Non-Linear Control using Neural Networks”
Proc. Control’92 Conference, Perth, 7-13.
Mills P. M., Wauchope M. D. and Burke P. D. (1992)
“Supervisory Control of the HIsmelt* Process”
Proc. Control’92 Conference, Perth, 7-13.
Mills P. M., Zomaya A. Y. and Tadé M. O. (1993)
“Applications of Adaptive Neural Model-Based Control”
32nd Conf. on Decision and Control 1993.
Burke P. D., Mills P. M. and Walker G. J. (1994)
“Model-based Control of the HIsmelt* Process”
Proceedings of Aspenworld 94 Conference, Boston, Ma. USA.
Duffy, G., Mills, P., Li, Q., Vlacic, V. (2015) “On the Industrial Plant
Performance & Operating Point Drifting Phenomenon.” In Proc. 10th
Asian Control Conference on Emerging Control Technique for a Sustainable World, Kota Kinabalu, Sabah, pp. 2187-2192.
Duffy, G., Mills, P., Li, Q., Vlacic, V. (2016) “Extending the Harris Index Performance Assessment Technique: A Plant-Wide Focus”, In Proceedings from the 7th Australian Control Conference, Newcastle, Australia.
Duffy, G., Mills, P., Li, Q., Vlacic, V. (2017) “A methodology to determine the dynamic relationship between process and manipulated variables”, In Proceedings from the 8th Australian Control Conference, Gold Coast, Australia.
Duffy, G., Mills, P., Li, Q., Vlacic, V. (2017) “An on-line process dead-time estimation algorithm”, In Proceedings from the 11th Asian Control Conference, Gold Coast, Australia.
Duffy, G. (2003) “Introduction to Quantum Computing Concepts”.
I N N O V A T I O N [ X ] is an engineering consultancy providing diagnostic, design and implementation services to industry, optimising existing mineral processing assets. The approach is to ‘challenge normal’, creating solutions bespoke to the customer’s specific needs. We have an award winning proven track record, generating significant bottom line cash flow benefit for our clients.