Computed tomography is a powerful imaging technique, which has in the mid-1970s revolutionized medical diagnostics as it gives doctors insight into the human body. Later, it has also found wide application in non-destructive testing. The idea to use computed tomography for monitoring industrial processes has emerged in the 1990s and coined the term ‘process tomography’. Since then, different process tomography modalities have been brought to live.
In medicine we commonly find X-ray tomography, positron emission tomography and magnetic resonance imaging. Medical tomography scanners are, however, complex, bulky and costly devices and need protection measures against ionizing radiation or strong magnetic fields. Process tomography comes with rather smart measurement systems and uses other signal carriers. Most prominent are
Electrical resistance (ERT) and capacitance (ECT) tomography measuring the conductivity or permittivity distributions within an object from electrical boundary measurements ;
Wire-mesh tomography (WMS) measuring the same in gas-liquid flows in a direct way via an electrode mesh ;
Magnetic induction tomography (MIT) measuring resistivity distributions from ac current excitations via boundary coils ;
Inductive flow tomography (IFT) measuring velocity distributions in conducting liquids from multiple induced voltage measurements at the vessel boundary ;
Contactless inductive flow tomography (CIFT) measuring 3D velocity fields in liquid metals from remotely measured disturbances of an applied magnetic field pattern ;
Microwave tomography (MWT) measuring dielectric loss distributions from multiple microwave transmission patterns ; and
Ultrasound tomography (UST) measuring acoustic impedance distributions from multiple sound transmission paths .
A reason for the existence of so many different modalities is their different capabilities to obtain certain process parameters. With respect to technical maturity it can be stated that ERT, ECT and WMS have achieved TRL 7 to TRL 9, while more recently developed MIT, IFT, CIFT, MWT, UST have TRL 6 or lower. However, it needs to be considered that process control with tomographic sensors has yet been demonstrated with simulations only. For this and also for harsh process conditions (high pressures and temperatures, fouling problems, strong electromagnetic interference) TRL is often below 3.
As soft-field modalities are mainly non-linear and ill-posed, linearized one-step image reconstruction approaches with regularization are normally utilized. With increasing computational power more accurate algebraic reconstruction schemes, like non-linear iterative reconstruction, become applicable both in 2D and 3D . A-priori process knowledge can make image reconstruction more robust against incomplete or corrupted data. Automated data analysis has always been an essential part of tomography-based process diagnostics. A fundamental step is mapping the reconstructed field quantities, e.g. conductivity or permittivity (ECT, ERT, MIT, WMS), acoustic impedance (UST), dielectric loss (MWT), current density (IFT, CIFT), to physical process parameters (phase holdup, density, temperature, moisture, species concentration). More advanced analysis then deals with feature extraction. Examples are bubble size measurement, interface detection or mixing time estimation.
Electrode-based ECT, ERT and WMS process tomography sensors will be qualified for harsh industrial environments via new materials and manufacturing technologies, such as tailored sensor ceramics and carbon-fibre reinforced sensor carriers with integrated functionality (HZDR). To achieve fouling resistance in ERT and UST, robust heated electrode and transducer designs will be devised. For CIFT and MIT magnetic sensors for extreme temperatures and sensitivity to magnetic fields several orders magnitude smaller than the applied magnetic field will be developed. This includes gradient coils with special mounting technologies, EMC compatible designs and new low-frequency ac excitation schemes. State of the art electronic hardware designs are used to securely detect micro-volts, pico-farads, and nano-amperes. Further, combined sensing-actuation strategies will be developed and demonstrated for the first time for MWT and UST. As electromagnetic probing is a commonality of the different modalities in TOMOCON this creates synergies for the development of sensor electronics, field simulations and data acquisition protocols and fosters collaboration among the ESRs.
With respect to data processing TOMOCON will focus on massive parallel real-time computation, new efficient algorithms for feature extraction (e.g. level-set methods, optical flow methods), and integration of tomographic sensors into a networked environment (e.g. data bus concepts and standards). The latter includes ways into standardization of data structures, data reduction and data transmission protocols for massive tomographic data. Research will address questions of sensors adaptivity, fault-tolerance and resilience. Another research direction is direct derivation of control data from multi-dimensional tomographic sensor raw data instead of transformed image data. As this is at a high abstraction level, new concepts of control theory and extensive use of advanced mathematical analysis concepts are required. At the far end the Internet of Things will provide a connectivity tool between multiple tomographic sensor systems. A practical issue is the selection of appropriate scalable hardware architecture for machine-oriented massive data processing. GPU systems, like NVIDIA Maxwell Tegra X1 technology are short-listed here but also other technologies will be analysed and comparatively assessed.
 Process Tomography – Principles, Techniques and Applications, Authors: M S Beck Williams, Butterworth-Heinemann, 2012.
 M. J. Da Silva, E. Schleicher, U. Hampel (2007), Capacitance wire-mesh sensor for fast measurement of phase fraction distributions, Meas. Sci. Technol. 18, 2245-2251.
 L. Ma, M. Soleimani, et al., (2015). Experimental evaluation of conductive flow imaging using MIT. Int. J. Multiphase Flow 72, 198-209.
 O. Lehtikangas, M. Vauhkonen, et al. (2016): Reconstruction of velocity fields in electromagnetic flow tomography, Phil. T. Royal Soc. A. 374, 20150334.
 F. Stefani et al. (2004): Contactless inductive flow tomography, Phys. Rev. E 70, 056306.
 Z. Wu, (2015), Developing a microwave tomographic system for multiphase flow imaging: advances and challengesT. Inst. Meas. Control, 37, 760–768.
 M. Yang, H. I. Schlaberg, B. S. Hoyle, M. S. Beck, C. Lenn (1997), Parallel Image Reconstruction in Real-Time Ultrasound Process Tomography for Two-phased Flow Measurements, Real-Time Imaging 4, 295-303.
 R. Banasiak, R. Wajman, D. Sankowski, M. Soleimani, (2010): Three-dimensional nonlinear inversion of electrical capacitance tomography data using a complete sensor model. Progr. Electromagn. Res. 100, 219-234.