The primary goal of this research is the creation of an ``intelligent and self-optimizing, adaptive, model-predictive'' (ISOAMPC) control methodology, which can predict process trajectories into the future, using the past implemented control actions plus a step or impulse process model and proposed future control actions, and then optimize performance in some desired sense. The novel characteristics of the extended method under development involves online model and process trajectory refinements using an infinite prediction horizon approach, and the application of constrained optimization for evaluating an optimal set of control moves. Current applications of predictive control include petroleum refining [1] and petrochemical processing [2].
One intermediate-term objective of research in this area is the development of an intelligent sensor for monitoring the temperature of molten plastic within a mold cavity as it cools. So far, this monitoring has eluded researchers, because sensor contact only provides surface measurements, so it is very difficult to measure the spatial part temperature distribution as it cools. It is envisioned that an intelligent sensor with the attributes of noise rejection, fault diagnostics, onboard active filtering and model-based estimation would significantly improve temperature distribution sensing and thus reduce part warpage and optimize cooling duration. For high volume production, reduction in part cooling time was obtained using conventional sensors and an energy conservative heat conduction model [3]. A further significant reduction in part cooling time would result using the proposed intelligent sensor and an improved heat transfer model. In the longer term, such approaches and technologies will be applied to other applications of interest to process industries.
The primary goal of this research area is to create highly autonomous sensors that exhibit intelligent and predictable behaviour, by developing modular systems that can be added to standard sensors. Results from this research will provide a key component in systems where failures or shutdowns cannot be tolerated. This is typical in assembly operations, navigational operations, industrial processes (e.g., injection molding), robotics, and aerospace applications. New research [4] will extend the investigations already conducted on Intelligent Sensors [5, 6]. Key features of the highly autonomous sensor (HAS) include the knowledge domains associated with the sensor and the measurand, qualitative interpretation of measured data, learning capabilities, and the adaptability to dynamic environments. These features will allow the HAS to reject electronic or mechanical interference, provide sensor and instrumentation operating parameters and characteristics, automatically correct for changes in the environment, and provide self-checking at startup/shutdown and periodically during operation. This would simplify and enhance the performance of manufacturing processes and other applications where complex control is required and sensor failures lead to expensive loss in production or production of substandard output.
Work in this area was initiated in 1996 [7], and progressed through a number of extensions and refinements [8, 9]. The overall objective of this effort is to develop control systems for plastics processing that produce higher quality plastic parts with minimum rejects. The activity in this area is currently focussing on online process model identification for different plastic materials, and cost function constrained optimization. Further research will input the plastic part dimensional characteristics for optimizing the process requirements and facilitate online learning and adaptability. The new ISOAMPC is to be developed for Ropak Canada Inc.\ located in Springhill, Nova Scotia, a high volume plastic production plant, in collaboration with the National Research Council, Boucherville, Quebec.
Research Assistant: Ms. Janet Beyea (commenced January, 1999)
- Cutler, C.R. and Hawkins, R.B. (1987), ``Constrained Multivariable Control of a Hydrocarbon Reactor'', Proceedings of the American Control Conf., p. 1014.
- Prett, D.M. and Gillette, R.D. (1980), ``Optimization and Constrained Multi-variable Control of a Catalytic Cracking Unit'', Proceedings of the Joint Automatic Control Conf., Paper WP5-C.
- Dubay, R., and Bell, A. C. (1998), ``An Experimental Comparison of Cooling Time for Cylindrical Plastics Components using Heat Conduction Models in the Non-Conservative and Conservative forms'', Polymer Engineering and Science Journal, Vol. 38, No. 7, p. 1048.
- Beyea J., Dubay R., and Bendrich G. (1998), ``Design of a data acquisition system for predictive control of melt temperature'', Society of Plastics Engineers ANTEC. (under review)
- Figueroa, F., and Egui, P.P. (1993), ``Development of a generic model of an intelligent sensor using the object oriented paradigm'', Proceedings of the 12th IFAC World Congress, Vol. 9, p. 183.
- Figueroa, F., and Mahajan, A. (1994), ``Generic Model of an Autonomous Sensor'', Mechatronics, Vol. 4, p. 3.
- Dubay, R. (1996), PhD. dissertation, The Utilization of Model Predictive Control for a Plastic Injection Molding Machine, DalTech, Dalhousie University, Nova Scotia.
- Dubay, R., Gupta, Y.P., and Bell, A. C. (1998), ``Predictive Control of Plastic Melt in Heated Zone with Insulation'', Journal of Injection Molding Technology, Vol. 2, No.1, p. 37.
- Dubay, R., Bell, A. C. and Gupta, Y.P. (1997), ``Control of Plastic Melt Temperature: A Multiple Input Multiple Output Model Predictive Approach'', Polymer Engineering and Science Journal, Vol. 37, p. 1550.
Information supplied by: Rickey Dubay
Last update: 1999 January 10
Email questions/comments/suggestions to: Rickey Dubay (dubayr@unb.ca)