An intelligent control strategy for container filling operations

Tools

Jeffries, M., 2000. An intelligent control strategy for container filling operations. PhD, Nottingham Trent University.

Preview

Text
10183228.pdf - Published version
Download (31MB) | Preview

Official URL: http://gateway.proquest.com/openurl?url_ver=Z39.88...

Abstract

As the need to reduce waste becomes an increasingly dominant force in the modern manufacturing world, the search for a reliable and accurate method to control container filling has been brought to the fore. Although modem bottling plants are sophisticated in mechanical design, the employed principles no longer provide sufficient scope for the optimisation necessary to meet waste reduction targets.

The use of ultrasound as the primary sensor for bottle filling plant has a great potential to provide the information on the process without significant redesign of mechanical systems or operational methods. However, the signals are strongly influenced by the properties of the fluids through which they pass - factors such as CO2, thermals and turbulence - leading to a randomising of the measured variable and thus the measurement process is classed as stochastic.

The presence of measurement noise can undermine the efforts of control systems to provide robust control. However, estimation theory in the form of a Kalman filter has the potential to overcome this obstacle, whether the noise is distributed in a Gaussian or non-Gaussian manner. Furthermore, the Kalman filter is a means to permit the use of ultrasound measurement techniques within the disruptive environment of the bottle filling process and allows an intelligent control strategy to be implemented.

A simple test rig is outlined in its use as a means of obtaining information on the characteristics of level measurement. The noise is categorised and a gamma distribution is shown to provide a good approximation to the noise statistics. From these results a single vessel simulation is used to investigate the performance of the proposed Kalman Filter under a number of different operational scenarios. It is shown that under some conditions, like changes in valve closure characteristics, the filter does not provide an adequate solution alone and that in some cases the controller must also be modified.

Fuzzy logic has many benefits for the design of a bottle filling control strategy. It offers good levels of robustness and stability without necessarily having the full knowledge of the system dynamics by using heuristic conjecture. In order to provide an optimal solution, the mathematical transformations that represent the fuzzy process have been investigated. To this end several avenues of modifications have been offered to better represent the needs of the filling process. These have focused on the input-output relationship, allowing a 'Fuzzy Flow Estimator' system to be developed, capable of parallel sampling ultrasonic Doppler shifts, and thus intelligently inferring both individual and global flow rate changes. This information is then used to feedback into the Kalman filters of the individual valve controllers.

Using an expanded version of the single valve plant simulation, several examples are shown on the performance of the Fuzzy Flow Estimator within a multiple valve carousel. The simulation results indicate that the proposed approach offers useful characteristics that would be beneficial for the tracking of both local and global disturbances in flow rate. Furthermore, these flow rates can be exploited by a Kalman filter to improve height or volume estimation at the individual valves, which provides increased robustness and a solid foundation for liquid level control in a bottling environment.

Item Type:

Thesis

Creators:

Jeffries, M.

Date:

2000

ISBN:

9781369314946

Identifiers: