Spatial modelling for electrical capacitance tomography

 


This interdisciplinary project is funded under the Computational Engineering Mathematics initiative of the EPSRC. The project will investigate the use of statistical spatial and temporal modelling for dynamic imaging.

 

The project research group are:

.     Dr Robert Aykroyd (Department of Statistics)

.     Dr Sha Meng (Research Fellow, Department of Statistics)

.     Dr Robert West (Biostatistics Unit)

.     Professor Richard Williams (SPEME)

.     Andrzej Romanowski (Visitor from Lodz, Poland)

.     Krzysztof Grudzien  (Visitor from Lodz, Poland)

.     Brian Cattle (PhD student in Applied Maths, Leeds)

.     Andrew Hillier and Fayyaz Rashid (Summer students from Leeds University, Mathematics)

Dr Bill Lionheart (Department of Applied Mathematics, UMIST) is collaborating on the project. Also thanks to Dr Mi Wang (CPACE, University of Leeds), Dr Manuch Soleimani and Dr Nick Polydorides (Department of Applied Mathematics, UMIST).

Andrzej and Krzysztof on the approach to Malham Cove (Spring 2004).

Andrzej and Krzysztof after the accent of Helvellyn (Summer 2004)

 


Project Outline

Electrical tomography techniques have many potential applications to industrial processes. We have targeted electrical capacitance tomography (ECT) for on-line monitoring and control of solids. The engineering aspects provide achievable goals with great cost benefit. Success requires the solution of mathematically challenging problems and interaction of mathematicians and engineers.

Currently linear approximations are used to produce process visualisations, which have artefacts including blurring, masking, shadowing and distortions. Few methods exist to process or analyse such images to enable automatic monitoring or control. The work here includes the development of statistical methods for nonlinear reconstruction that minimise artefacts to improve visualisation and allow direct estimation of control parameters. This requires extension of existing statistical image analysis. The proposed MCMC estimation produces samples of images quantifying confidence and identifying alternative plausible scenes. For high-level modelling, the estimation produces a distribution of control parameters assisting the formulation of control strategies. Real-time control requires temporal extensions and greater efficiency of sampling. The importance of parameters will be ranked by the posterior distribution and an adaptive sampling strategy proposed both for the measurements of the industrial process and the MCMC sampling.


Research Outputs:

Papers and Reports:

Grudzien K, Romanowski A, Sankowski D, Aykroyd RG, Williams RA (2005). Advanced statistical Computing for Capacitance Tomography as a Monitoring and Control Tool.  5th International Conference on Intelligent Systems Design and Applications.

West RM, Meng S, Aykroyd RG and Williams RA (2005). Spatial-temporal modelling for EIT of a mixing process. Full paper To appear in Review of Scientific Instruments.

Grudzien K, Romanowski A and Williams RA. (2004). Application of a Bayesian approach to the tomographic analysis of hopper flow. In submission.

Romanowski A, Grudzien K, Aykroyd RG and Sankowski D (2004) Application of the Bayesian/MCMC approach to the tomographic investigation of multiphase processes. In submission.

Aykroyd RG, West RM, Meng S and Williams RA (2004).  (Contribution - html), Contribution to the Discussion at the RSS Ordinary Meeting on Inverse Problems.

West RM, Aykroyd RG, Meng S and Williams RA (2004). MCMC techniques and spatial temporal modelling for medical EIT.  Physiological Measurement, 25, 181-194.

Meng S, Aykroyd RG, West RM and Williams RA (2003). Geometric Modelling for Medical EIT. In R.G. Aykroyd, K.V. Mardia and M.J. Langdon (Eds.), Stochastic Geometry, Biological Structure and Images, pp. 101-104. Department of Statistics, University of Leeds.

Aykroyd R.G, Meng S and West R M (2003). Spatial-temporal modelling for a nonlinear inverse problem in industrial process tomography. Research Report No. STAT-03/05.

Meng S, Aykroyd RG and West RM (2002). An investigation of deterministic and stochastic strategies for the solution of inverse problems. Research Report No. STAT-02/08.

Aykroyd RG, Meng S and West RM (2002). Deterministic and stochastic approaches to inverse problems. In RG Aykroyd, KV Mardia and P McDonnell (Eds.), Statistics of Large Dataset: Functional and image data, bioinformatics and data mining, pp. 85-88. Department of Statistics, University of Leeds.

West RM, Aykroyd RG and Meng S (2002). Markov chain Monte Carlo techniques for inverse problems. Research Report No. STAT-02/05 .

 

Seminars and Conference Presentations:

 

4th World Congress on Industrial Process Tomography, Japan, September 2005:

Modelling and predicting flow regimes using wavelet representations (DA Goodwin, RG Aykroyd, and  S Barber), Vol 2, pp 904-909.

PostScript; PDF

A boundary element formulation for object tracking from ERT data (RG Aykroyd, BA Cattle and RM West), Vol , pp 637-642

PostScript; PDF

Full shape reconstruction from partial ERT monotonicity information using a Bayes-MCMC approach (RG Aykroyd­, M Soleimani and WRB Lionheart), Vol 2, pp 697-702.

PostScript; PDF

MCMC algorithm acceleration using a hybrid linearised/non-linear forward problem strategy applied to Bayesian analysis of 3d ERT (RM West, M Soleimani, RG Aykroyd­, and WRB Lionheart), Vol 2, pp570-575.

 

Multi-modal data fusion for enhanced imaging applied to 3D ERT with ultrasound time of flight (Soleimani, Aykroyd, Freer, Lionheart, Podd), Vol 2, pp663-668.

 

A novel approach to pneumatic conveying monitoring and control strategy development (Grudzien, Romanowski, Aykroyd and Williams), Vol 2, pp 886-891.

 

5th International Conference on Inverse Problems in Engineering: Theory and Practice, Cambridge July 2005:

Boundary element method and Markov chain Monte Carlo for object location in electrical impedance tomography (Robert G Aykroyd and Brian A Cattle)

PostScript; PDF

 

Markov chain Monte Carlo strategies applied to nonlinear inverse problems  (RG Aykroyd, BA Cattle and RM West) To be presented at the BMVA meeting on Optimisation, London November 2004.

Object tracking from electrical tomography data using Boundary element methods (by RG Aykroyd, BA Cattle and RMWest) Presented at the BMVA meeting on Image Features and Statistics, London October 2004.

Bayes-MCMC reconstruction from ERT data with prior constraints from resistance matrix monotonicity. (by RG Aykroyd­, M Soleimani and WRB Lionheart) Presented at the 3rd International Symposium on Process Tomography in Poland.

An introduction to the Bayesian/MCMC approach as a recently developed methodology for tomographic data analysis (by Andrzej Romanowski, Krzysztof Grudzien, Robert G Aykroyd, Richard A Williams, Robert M West and Sha Meng) Presented at the 3rd International Symposium on Process Tomography in Poland.

Application of the Bayesian approach to powder flow investigations (by Krzysztof Grudzien, Andrzej Romanowski, Richard A Williams, Robert G Aykroyd, Robert M West and Sha Meng) Presented at the 3rd International Symposium on Process Tomography in Poland

Boundary element methods and Markov chain Monte Carlo for fast object location applied to electrical tomography problems (by Robert G Aykroyd, Brian A Cattle and Robert M West) Presented at the 3rd International Symposium on Process Tomography in Poland

Bayes-MCMC reconstruction from 3D EIT data using a combined linear and non-linear forward problem solution strategy (by M Soleimani, RG Aykroyd­, RM West, S Meng, WRB Lionheart and N Polydorides). Presented at the 5th International Conference On Electrical Bio-Impedance Electrical Impedance Tomography, Gdansk, Poland, 2004. [An extended abstract is available.]

Estimation of monitoring and control parameters of industrial processes from ECT data: The Bayesian/MCMC approach (by Andrzej Romanowski and Krzysztof Grudzien) presenting at the VCIPT Meeting 1-2 April 2004 [Abstract]

Knowledge-based modelling for dynamic electrical tomography in medicine and process engineering (with RM West, S Meng and RA Williams). (Poster abstract - html) Presented as a poster at the BMVA Symposium on Spatiotemporal Image Processing, March 2004

Knowledge-based modelling for electrical tomography in medicine and industrial process engineering (Abstract - html)

University of Edinburgh, January 2004

Spatial-temporal modelling for EIT of a mixing process (jointly by RM West and RG Aykroyd)

3rd World Congress on Industrial Process Tomography, Banff, Canada, September 2003.

Geometric modelling for electrical impedance tomography (by RG Aykroyd)

The 22nd Leeds Annual Statistical Research Workshop. July 2003.

A Bayesian approach to inverse problems in medicine, archaeology and industrial process engineering (by RG Aykroyd) (Abstract - html) 

St Andrews University / RSS Highlands Local Group, May 2003

University of Sheffield, February 2003.

MCMC techniques and spatial temporal modelling for medical EIT (by Sha Meng)

4th Conference on Biomedical Applications of Electrical Impedance Tomography, UMIST, April 2003.

Spatial-temporal modelling for tomography with MCMC techniques (Poster)

Isaac Newton Institute, Cambridge University, January 2003.

Bayesian modelling for industrial process tomography (by RG Aykroyd) (Abstract - html)

Northern RSS Local Group, Newcastle University, December 2002.

Deterministic and stochastic approaches to inverse problems (by RM West)

The 21st Leeds Annual Statistical Research Workshop. July 2002.

MCMC techniques for inverse problems (by RM West)

British Inverse Problems Workshop, Leeds, April 2002.

 


Other relevant links:

Department of Mining & Mineral Engineering

3rd World Congress on Industrial Process Tomography

Virtual Centre for Industrial Process Tomography

XII International Conference on Electrical Bio-Impedance, Gdansk

Centre for Computational fluid dynamics

3rd International Symposium on Process Tomography in Poland

Non-Newtonian Fluid Dynamics

4th World Congress on Industrial Process Tomography

 


Back to Robert Aykroyd's Homepage