Smelling a fat

We reported on electronic noses two years ago, but this sniffing technology has moved on BY CHRIS WEBB

Rancidity is a subject likely to raise a few eyebrows – and pucker a few noses – at the Leatherhead Food Research Association (LFRA). Researchers working for its members’ Edible Oils and Fats Panel want to check the state of health of the product and, what’s more, predict its shelf-life at a particular storage temperature. And the latest weapon in its sniffing arsenal? Why, electronic noses, of course.

Ultimately it is hoped a single machine could complement but significantly reduce the need for the hugely costly sensory panels which typically require up to 12 trained analysts at a time.

Although the technology has been around for about three years, only now are the implications of its availability beginning to dawn. Indications are that food freshness and packaging taint detection are likely applications, while authenticity testing – of products such as coffee, for example – is also a possibility. There has even been talk of wine testing.

Dr Vince Shiers at the LFRA has to date investigated the potential of a number of instruments, among them Alpha MOS, FOX 3000, AromaScan, A32S and Nordic Sensor Technology.

The results from these new instruments are compared against an accepted benchmark for the more qualitative results obtained by sensory panels, microbiological and chemical determinations.

Says Shiers: ‘Identification of standard materials that might be sufficiently representative of the samples under test, while still being stable to any change over time, should enable repeatability, reproducibility and any sensor drift to be detected.’

Findings so far indicate that at least one, and perhaps more machines are capable of detecting several early and intermediate stages of oil rancidity.

On a technological level, it is perhaps one of the most exciting developments in the food and beverage industry – where automation is now in reach of even small to medium size producers.

This is a point underlined by ABB’s John Croft, the company’s manager for Manufacturing Execution Systems (MES). ‘The last year or year-and-a-half have certainly made it easier for smaller companies to take advantage of the technology formerly only within reach of much larger, bulk producers. For example, in batch processing using smaller and relatively cheap PC based systems.’

Croft likens the move to Just In Time production methods now widely used in other areas of manufacturing, a move that has reduced dead stock and boosted efficiency.

Quick reactions

‘Typically, the food industry will turn out many products to short delivery times for a number of major consumers. The new technology enables finite capacity scheduling – based on command, not forecast – to make the most of resources.’

Other benefits have come into range with it: automation has made available the possibility of a more open environment, bringing into play numerous information management facilities, meaning integration and benefits at both ends of the manufacturing cycle – not just purchasing, but also supply.

Ken Adamson, regional sales manager for Alfa Laval with a specific responsibility for the company’s food and beverage activities, says batch processing has brought with it significant developments in software, rather than hardware.

‘As systems become more open, specialist functions become more widely available to all platforms – batch control sytstems, warehousing, and communications protocols such as Profibus and Fieldbus. The dilemma facing the end user is what system should he invest in as the distinction between DCS and PLC/SCADA systems diminish with the advent of standard software packages such as MS Windows.’

Traditionally, says Adamson, the approach in the food industry has been to mount all the I/O together with the CPU. ‘But it’s not very economical. It’s better to install a flexible hardware platform which gives the ability to accommodate central or remote I/O.’

Alfa Laval’s PC based SCADA system is its modular SattGraph 5000. Developed for the 32 bit Windows NT platform, the open system architecture is based on established standards like Open Data Base Connectivity (ODBC), OLE, DDE and MMS.

Communication between the operator stations and the PLCs is via Ethernet. But this approach does have its drawbacks, not the least being that the PLC/PC hybrid implies seperate training and experience sets, not to mention two programs that might have to be supplied by two different software manufacturers.

The resulting possibilities are object linking of graphics and animation, operator controls, control logic, alarms, recipes and data logging, etc. ‘In the case of a dairy, for example, this can mean that modules are in turn grouped together in larger objects: filling lines, reception, CIP, pasteurisers, etc.’, says Adamson.

Different application libraries will be used for dairies, breweries, juice plants or those producing carbonated soft drinks.

The brewing industry has, of course, been one of those at the forefront in the sense of pushing forward the automation expectations of food and beverages.

Traditionally, a brewery could have been dedicated to producing only one, or at most a few types of beer, whereas today’s brewery may have to produce for many different brand names. Running one type of beer for as long as possible minimizes changeover costs, but increases the possibility that storage capacity will be exceeded, and/or the inventory levels of other products may reach unacceptably low values.

Furthermore, the brewery operator may be required to respond at any time to changes in production requirements, the state of brewery equipment, etc.

Advanced software techniques – Genetic Algorithms implemented within an object-orientated, expert system environment – have proved valuable when applied to this type of problem.

The Genetic Algorithm approach has many advantages over traditional optimization techniques.

* It is very flexible, in that fitness can be measured as needed, including highly customized and unconventional definitions of objectives and constraints.

* It can be fast when there is a need to sample a large solution space, where there are many combinations of alternatives to consider.