On Line Instrumentation

4.1. Design considerations

In a few cases moisture measurements have been carried out on line, or at least beside a production line. In order to build an on line system a number of special spatial and temporal considerations have to be considered (Bjorkstam and Listerud, 1985). Most of these relate to the design of the NMR magnet and the radio frequency excitation coil. The sample must pass through both components simultaneously. Both must clearly be sufficiently large to accommodate the product but an overlarge magnet is unduly expensive and an overlarge excitation coil leads to excessive radio frequency power requirements and, due to a poor filling factor, poor signal to noise. The design is severely constrained by a basic rule of NMR which states that the static and excitation magnetic fields must be orthogonal. Therefore, whilst a simple approach is to pass the production line axially through a solenoidal excitation coil it is not possible to use a solenoidal electromagnet for the static field as well. However, the production line may be passed between the poles of an iron yoke double E electromagnet or of a permanent magnet whilst using a solenoidal excitation coil. If this is not possible, it may be necessary to use other excitation coil geometries such as birdcage resonators or surface coils, both of which are well established for medical imaging applications. Birdcage resonators can be constructed around a cylinder with the excitation field parallel to one diameter. Generally, it is found that the excitation field is more uniform although a little weaker and the tuning more stable in birdcage resonators compared to similarly sized solenoids. Surface coils are flat and can be laid over the sample. They suffer from the disadvantage of small excitation volumes. So called one sided magnets which can be placed below a production line are also available but these suffer from the disadvantage that the homogeneous field volume is small and generally flat. It may not be representative of the whole sample if, for instance, the product settles and stratifies on a conveyor belt. Magnetic field strengths used in on line applications are generally of order 0.25 T corresponding to a hydrogen proton resonance frequency of just over 10 MHz. This limit is brought about by size and cost of the magnet and generally decreases with increasing sample volume.

A major design and cost factor relates to radio frequency power. In order to observe the short T2 components of the FID, essential for a solids determination, it is necessary to have a very short and very intense excitation pulse. Apart from having large and expensive radio frequency amplifiers (and powers in excess of 1 MW have been reported (De Los Santos, 1994)), intense radio frequency pulses can be obtained by using highly tuned (i.e., high Q) excitation coils. However, equally necessary is a rapid recovery of the receiver system following the pulse and in particular a short ring down time for the coil. This generally requires lower Q. One option is to incorporate Q switching and active damping technology, another is to use separate receive and transmit coils (Fukushima and Roeder, 1981). Increasing the NMR frequency helps and also dramatically improves the signal to noise ratio of the experiment. However, the cost of the magnet can increase unacceptably fast. As a guide, for a fast recovery system with a sample approximately the size of a loaf of bread the required transmitter power will be of order 10-100 kW.

The required length of the magnet is determined by the speed of the production line and by the relaxation times Tj and T2. For a sample with a spin lattice relaxation time of 1.0 s and a production line moving at 0.2 m/s, a polarising magnet 0.6 m long is needed in order to produce an initial magnetisation which is 95% (three decay constants) of maximum. If the sample T2 is 0.5 s then in order to see two decay constants of the FID the subsequent measurement magnet will need to be a further 0.2 m long. To this must be added the length of the product, say 0.2 m, giving a total length of 1 m. One large magnet may be used. If two separate magnets are used, the first can be of much lower homogeneity and can be set at a higher field strength so as to increase the magnetisation polarisation and hence the signal to noise ratio of the subsequent measurement. The second needs to be of significantly greater homogeneity. Permanent magnets of this size, whilst relatively inexpensive, are notoriously heavy. Electromagnets can give greater field strengths but considerations of electrical power consumption and physical bulk tend to favour the permanent magnets. If the signal to noise ratio of the measurement permits, and only reproducible signal amplitudes are of interest, then it is possible to use shorter magnets. For instance, if 50% of the equilibrium magnetisation provides sufficient signal then the polarization magnet in the above example can be reduced in length to just 0.14 m. Moreover, if it is not required to sample the magnetisation beyond 100 (is of the FID, then the measurement magnet need be barely longer than the product.


polarisation ^

3vT i measurement

Figure 3. A summary of the design features for an on line low resolution

NMR system around a conveyor carrying discrete products.

A number of the points discussed above are summarised in figure 3 which shows a schematic of a hypothetical low resolution NMR spectrometer around a conveyor carrying discrete products. Apart from the magnet and radio frequency coil, the main parts of an NMR spectrometer are the radio frequency receiver and transmitter and control computer. A small radio frequency amplifier is rack mountable, larger amplifiers are free standing. The remaining components are now usually housed in a desk top computer which also serves for data analysis and the output of feedback control signals to the process.

Environment presents another problem. Temperature stabilisation of the magnet is likely to be required in order to ensure field homogeneity and strength remain constant. This is more of a problem for magnets made from rare earth materials than iron but even these must be controlled in some way. Often this is done by having a frequency lock signal derived from a test sample in a small NMR coil near the main coil and sample in the magnet. The lock provides a signal for feedback control to magnet. The homogeneity of the magnet is spoilt by the introduction of extraneous ferrous material. Fixed installations can usually be shimmed out by the addition of a set of shim coils carrying carefully preset small electric currents. However, these will not cope with other ferrous objects brought near the system and a ferrous exclusion zone needs to be set up around the magnet. Depending on the size and design of the installation, this may be 1 m or more in radius. Workers in the vicinity of the apparatus may require screening for heart pacemakers and other metallic implants. Some commercially available magnets have substantially reduced the surrounding fringe magnetic field by the incorporation of active shielding. Actively shielded magnets can be placed almost anywhere in the production line. The radio frequency irradiation will not penetrate through metal. Thus, for instance, metal pipes containing flowing product must be replaced by non metal pipes in the vicinity of the sensor coil if the coil is to be external to the pipe. However, the presence of a metallic shield around the sensor coil dramatically improves signal to noise by reducing the influence of external electrical noise pick up and a sensor coil within the pipe may be advantageous, especially as it leads to an excellent filling factor. For hydrogen NMR, hydrogen rich conveyor or pipe material gives a background signal which must be calibrated out if the material cannot be changed. Glass and PTFE are often useful construction materials.

Technical and economic considerations often suggest that in line sampling may be more appropriate than on line analysis of the total product being produced. Sampling can take various forms of which two are particularly favoured. The first is a scheme which momentarily stops a fraction of the product, normally by removing it from the line. The second is a scheme which analyses a more slowly moving fraction of the product on a thief or side line. With both these options magnet size, radio frequency excitation power and general engineering difficulties associated with the NMR are reduced and so, therefore, is cost. Moreover, more time is available for the measurement leading to increased accuracy and sensitivity of the measurement. However, a suitable sampling procedure must be developed if the results of the measurement are to be extrapolated to the whole line. In some cases (e.g., foreign body detection) sampling is clearly impossible.

4.2. Experimental systems

Only a very few specific on line and large industrial applications of NMR have been developed and a subset of these have been documented in the literature and are in the public domain. They have been developed for a variety of applications not all of which are in the food industry. What many of them have in common, however, is the acceptance of effective sample characterisation according to an NMR derived signal without undue detailed interpretation of the information being obtained. As yet, no manufacturer offers a standard large scale on line system off the shelf, indeed to do so would be very difficult as production environments vary greatly. However, larger scale modular systems should become available in the foreseeable future and this should reduce costs. Two groups have made significant contributions. The first is based at Southwest Research Institute, San Antonio, Texas, USA. The second is associated with RM Pearson now at Tri Valley

Research, Pleasanton, California, USA. With both groups a number of systems have been tested and successfully used in the field.

The first published references to the use of NMR as a process control technique were made relatively early by Nelson of Varian Associates and Reilly and Savage of Shell Development Company in 1960 (Nelson et al, 1960; Nelson, 1964) and discussed a spectrometer constructed five years earlier and installed in a pilot plant in 1956. This was before the availability of pulsed techniques. The system used continuous wave NMR to monitor the chemical shift spectrum from a liquid stream with the entire spectrum being recorded once every 6 s. To overcome the problem of polarising the nuclei a small sample reservoir was included within the magnet immediately before the sensitive volume. The sample tube was 3 mm in diameter and the sample flow rate a few centimetres per second. Faster flow rates were not possible as the sample was not then in the magnet sufficiently long to prevent broadening of the high resolution lines. Analogue electronics was used to lock the system to one line in the spectrum and automatic gain control was used to keep it at constant amplitude. Thereafter, changes in a second line of the spectrum related to compositional changes in the liquid. The equipment worked at 30 MHz. More recently, Tellier and co-workers have modified a high resolution NMR sprectrometer to monitor the ratio of water and CH2 resonances in flowing fine meat pastes in order to determine the fat content (Tellier et al, 1990).

The development of a series of small nuclear magnetic resonance spectrometers for process control has been carried out by RM Pearson and colleagues, first at Kaiser Aluminium and more recently at Tri Valley Research (Pearson and Job, 1992; Pearson et al, 1987; Pearson and Parker, 1984). Initially, three small Bruker / IBM bench top systems were adapted for installation on line in Kaiser Aluminium oxide plants. They were able to satisfactorily measure the moisture in aluminium oxide on line although the instruments were far from user friendly. Subsequent spectrometers were better. A major problem for these systems was temperature stabilisation. Working at Tri Valley and with the aid of more modern computing systems and more advanced magnet technology a range of systems has been developed including one for measuring moisture and oil in cereals. As with the other applications already discussed, the systems are small and based on sampling a side stream of the main production line. The latest use Halbach magnets which are cubic with a side of approximately 225 cm and a sample access tube of over 3 cm diameter and are being developed for process control in a hot asphalt mix plant.

Much of the work at Southwest Research Institute is reviewed in a paper by Nicholls and De Los Santos (Nicholls and De Los Santos, 1991). They describe a system which encloses a side stream of the production line, a system based on a magnet which resides below the production line using one sided magnet and coil geometries and a system which resides below the line but samples from it into a more conventionally shaped spectrometer via a piston. A schematic, reproduced from the original paper, of this last system which has been built and successfully tested in the field, is shown in figure 4. It was developed under a US Department of Energy contract as a means for improving energy conservation in agricultural drying plants. A mechanical piston is lowered and product from the line falls into a sample cell where it is measured. After the measurement the sample is returned to the line by the piston and a fresh sample taken. It is suggested that this has several advantages, notably the compactness of the spectrometer and the minimal changes required to the production line (drilling a hole in the bottom). No side streaming is required. The procedure is particularly suitable for dry products. It is possible to build standard reference samples into the piston which can be measured as the piston rises / falls so that temperature drifts and other extraneous problems are greatly alleviated. The one sided system they describe suffers from the disadvantage that only a small volume of the product is sampled and that it is always taken from the same area of the production line. The authors estimate that the cost of a sensor is of order $30-50,000 (1991) and that the payback period in a large gluten drying plant could be of order 1 year.

Figure 4. A schematic diagram of a piston sampling LR NMR sensor. The design is the subject of US patent No. 5129267,1992. (adapted from Nicholls and De Los Santos, 1991)

It is worth mentioning, if only for curiosity, that workers at Southwest Research Institute have also developed an NMR sensor which can be transported on the back of a tractor (Paetzold RF et al, 1985). This machine has been successfully used to measure sub surface soil moisture at a depth of a few cm. Other large spectrometers developed at Southwest Research Institute include a baggage handling system designed for detecting narcotics and plastic explosives in suitcases. (De Los Santos, 1994).

Bore hole NMR logging tools (Kleinberg et al, 1992) are another excellent example of NMR spectrometers developed to work in extremely hostile

Relaxation Audio Sounds Autumn In The Forest

Relaxation Audio Sounds Autumn In The Forest

This is an audio all about guiding you to relaxation. This is a Relaxation Audio Sounds with sounds from Autumn In The Forest.

Get My Free MP3 Audio

Post a comment