This is adapted from CHAPTER 6 of the book 'Operations Research in Ireland' edited by Julian Mac Airt and published in 1988 by Mercier Press, Cork.

INTRODUCTION

The project work constituted the practical part of the master's degree programme; there was also a lecture programme and examinations. The work was innovative in three ways. First, each project involved more than one student, usually two or three. Secondly, the terms of reference were negotiated with an outside client or sponsor, who was interested in a real solution to a real problem. And thirdly, the supervisor (ie this author) was working as consultant to the client for a fee and was responsible for the delivery of the results, irrespective of the performance of the student groups.

This procedure was possible largely because the author had opted to develop his OR consultancy as a fringe-university activity, on the grounds of his belief in the need to develop the university post-graduate system as a useful resource.

The partition of the tasks among the members of the project team was an important supervisory activity. It constituted a problem in some cases where the matching of complementary skills within the group was less than perfect. Procedures were established for setting up teams and allocating them to projects during the first term of the academic year (October to December). Dry runs on textbook case studies gave some indication of student attributes; a voting procedure was established for enabling students to express preferences for one project or another. This was taken into account when the teams were set up in December.

The projects were worked on until the following September, with bursts of initial contacting and data-gathering in the Christmas and Easter vacations, followed by full-time activity after the theoretical part of the MSc course had culminated in the June examinations.

When setting up the teams, account was taken of students' primary qualifications as well as their preferences. A typical team of three would need to specialise: for example, one might concentrate on assembling and abstracting the available data into an appropriately structured (computer-accessible) database; a second might specialise in developing an appropriate decision-model capable of accessing the database for a series of 'what-if' projections; while a third, possibly playing in some ways a leading role, would determine the range of options to be tested in the light of the needs of the client.

These roles however always overlapped, and in all cases it was possible to examine each student in the project as a whole, drawing out in greater detail the work for which he or she had personally been responsible.

In some cases the student report was accepted by the client; in other cases the client-report was done by the supervisor (usually when the client's deadline was inconsistent with the academic schedule). In the latter case, the student report had the status of an interesting and possibly useful background study.

The clients were in all cases either state or parastatal agencies, or agricultural co-operatives. No work was done in this mode for private industry. This perhaps reflects the dichotomy of the Irish private sector, which subdivides into (a) transnational corporations, which have their own in-house R and D facilities, usually abroad, and (b) small to medium-sized Irish-owned firms, which tend to be conservative rather than innovatory. It would appear, therefore, that there is a positive role for the state and the agricultural co-operatives in harnessing the university research system to the needs of national development.

This discussion first gives brief abstracts of all 7 projects. The first of these presents some interesting features (it is to do with the re-introduction of flax to Ireland as an economic crop) and is dealt with in some depth. The next 3 projects, all concerned with milk, are discussed in somewhat less depth and act as a basis for re-stating the problem in the contemporary environment. Extracts are given from one other project involving the analysis of a relatively complex system - the rail network. Finally, 'the transition problem' is discussed - how the momentum of projects such as these can be kept up in the aftermath, in the real world, in cases where the concept has development potential. There are unresolved problems in this area.

PROJECT ABSTRACTS

Profitability of flax production in Ireland:

The linen industry in the north of Ireland has for some decades been dependent on imported flax, the decline in local production being attributable to problems arising in the old 'wet retting' process, which was labour-intensive and polluting. Technological changes in the industry have rendered 'dew retting' acceptable(2), so that the harvesting process now lends itself to modern agricultural mechanisation, involving investment in specialist machinery. The price is determined by quality, which depends critically on the retting process. The profitability of the overall enterprise depends on a compromise between level of mechanisation and losses due to delay in harvesting operations.

(1) This situation was modelled, by simulating the harvesting operations against a background of variable weather conditions, the latter determining both degree of retting and machine mobility. Information derived from the simulation was used to help determine the 1971 harvesting programme and in planning expansion for 1972.

Milk quality and bulk milk collection:

A simulation was developed to predict the overall bacterial quality of the milk in the tanker as a function of time, given the temperature and quality at successive pick-up points. The results of this contributed to the development of a procedure involving shift-work for tanker drivers, with scheduling to take account of the differing pump-off rates of refrigerated and unrefrigerated milk, collected separately. The penalty system for quality control was also tightened up. This plan was implemented and significant savings were achieved.

Milk supply seasonality in the processing industry:

Milk processing in Ireland is dominated by the fact that the supply tends to follow the seasonal growth of the grass: a summer-winter ratio of 12:1 or even 20:1 is considered normal. As a consequence, the bulk of the production is of storable products (butter, skim powder, hard cheeses, casein), sold mostly as bulk commodities.

A techno-economic model of the industry was developed which permitted its overall profitability to be examined in the context of an assumed reduction in seasonality ratio, (a) with the existing product mix and expanded throughput with existing capacity, and (b) with an enriched product mix (having a higher proportion of added-value products, such as would become possible with continuity of supply and improved quality and composition of milk). This expanded, deseasonalised production was compared with expanded production of the present mix, with capital investment to cope with the expanded peak capacity.

With the aid of this model, it was possible to estimate a winter premium price payable to achieve continuity of supply. The transition to the new situation, however, would be costly for the farmer (involving delaying the calving of a proportion if the herd for 6 months, with consequent loss of income) and would require positive management.

The presence of the EEC accession price bonanza militated against the recognition of the problem in 1972. It is appropriate to examine it again against the background of contemporary thinking on the CAP in the European Parliament and the existence of the 'superlevy'.

Milk seasonality and the farmer:

It was concluded - taking into account seasonal effects in the beef industry (which reflected into improved prices for off-season calves and 'canners') and the extra milk yield from the extended lactation of the autumn calver on the spring grass - that deseasonalising the milk supply would, in fact, break even. The cost to the farmer would, however, be appreciable since in the transition to autumn calving for a proportion of the herd revenue would be lost for a six-month period. This constitutes a significant barrier requiring financial incentive from the industry and/or the State if it is to be overcome.

Locomotive-linking in the national rail system:

A simulation of all passenger and freight train movement on the national rail network was developed, such as to enable the overall locomotive requirements to be estimated for a given schedule. It proved feasible to predict locomotive needs in accordance with current experience, given the existing schedule and constraints. This suggests that the technique might be used with some confidence to predict rapidly the locomotive requirements arising from revised schedules and/or revised constraints.

Computer-aided planning in second-level education:

By combining various derived measures with appropriate weighting factors (reflecting their perceived priorities), it was possible to define a 'figure of merit' which took into account (a) degree of satisfaction of demand for exotic subjects (such as the rarer foreign languages), (b) extent to which subjects were taught by teachers actually qualified in them, and (c) degree of teacher specialisation.

The model was partially validated with some survey data derived from a previous project which had analysed several existing schools in Galway with input to the present project in mind.

Glasshouse tomato production:

Output from this analysis was embodied in a predictor, giving days to first harvest under standard temperature conditions as a function of planting date. This was used as input to a more general-purpose planning model which enabled various production systems (cold-house, monocrop, duocrop) to be compared in a given market background.

PROFITABILITY OF FLAX PRODUCTION

The Bunclody Farmers' Co-op in Co Wexford, the client for this study, had been established in 1959; by 1970, it had 150 members and a turnover of £1m. Its main business was in the sale of wheat, barley, strawberries and honey on behalf of its members. There was a need for some other crop to rotate with wheat, in the interests of disease and pest control, with a comparable financial return. The first trials of flax were made in 1966 and in June 1968 the Co-op entered into a firm contract with Kirkpatricks Ltd of Ballyclare, Co Antrim, to harvest up to 3000 acres of flax. The 1970 planting was 610 acres. All was set for expansion of the production of an attractive crop.

In the expansion phase, it became important to plan the investment in, and operational procedures for, specialist machinery; this would be owned and managed centrally the Co-op. It was also important to be in a position to evaluate the crop from the point of view of the farmer. To achieve these objectives, two planning models were developed - the 'farm model' and the 'machinery model'.

The farm model enabled the costs associated with the cultivation and harvesting operations, and the revenue from the sale of the crops, to be calculated for a specific farm over a number of years, taking into account crop rotation effects. It was required to work either in historic mode or as a simulation for use in forward projection. It was non-optimising; it modelled simply the consequences of the farmers' decision regarding choice of crop and allocation of crop to specific areas of the farm. Crops catered for were grass, feeding and malting barley, oats, swedes, sugar beet, wheat and flax. There was provision for adjusting yields to respond to various rotation patterns. Thus the background of standard crops to which the flax was to be compared was reasonably credible.

Flax had to be treated in more detail, with the retting process quantified. The harvesting process involved pulling the flax up by the roots (it was this that requires the relatively expensive specialist machinery) and then leaving it to lie on the ground long enough for the soil bacteria to attack the cellulose, but not for so long that the fibre is attacked; this is the so-called 'dew-retting' process(2).

It was possible to define a 'retting index', to quantify the process and to relate it to a price, based on the tensile strength of the resulting linen. The relationship between the price and the retting index is determined by the market and was given as follows:

Rettinglndex:   0.1 0.2 0.3  0.4 0.5  0.6  0.7  0.8  0.9 

Price:          0    0  0.68 0.8 0.91 1.0 0.92  0.86 0.73

Thus one should aim for a retting index of 0.5 to 0.7; this has to he judged by field trials and subsequently verified by laboratory testing of the fibre.

In order to predict the cumulative retting index, estimates were made of the 'rate of retting' as a function of temperature and humidity. These were derived by combining the experienced judgement of Belgian and Irish experts with weather records. The judgement'is summarised in the following table:

RETTING RATE ADJUSTMENT MULTIPLIER

Weather Category     Weeks for Optimal Retting    Multiplier 
warm wet                         3                  2.00 
warm dry                        4-5                 1.34 
average                          6                  1.00 
cold wet                        7-8                 0.73 
cold dry                         9                  0.57 

The 'cumulative retting index' is obtained by multiplying the 'retting rate' by the 'multiplier' in each week, depending on the weather, and summing over the weeks to date.

The weather records were taken as weekly averages for the 17 weeks of 1 July to 31 October over 11 years (ie a sunspot cycle) and were allocated to the above categories. Retting rates were established for each of the 17 weeks under average conditions; these were temperature-dependent and ranged from 0.1034 for the beginning of July to 0.0503 for the end of October. The 'cumulative retting index' was then estimated for a given weather pattern as indicated above.

This retting index predictor, using weather records from Kilkenny (the nearest available to the area under consideration), gave good agreement with actual results on a sample of farms, so that when used in simulation mode in projections it had some credibility. Thus the farm model when run gave results which took into account the price receivable for the flax resulting from a given range of harvest scheduling and weather situations.

The second planning model devised - the 'machinery model' - looked at the harvesting process from the angle of the Co-op and considered the scheduling of movements of the two types of specialist machinery around the growing region. (As well as the puller, there was a binder to lift the retted flax and bale it for transport to the central scutching-mill.) As the period between pulling and binding is needed for the retting process, the timing of these operations is of critical importance.

Farms on the model were located on a grid of 48 zones covering the whole region. A machine when finished on one farm moves to another in the nearest neighbouring zone which is known to be 'ready'. A farm has four states: not ready, ready, being worked and finished. A machine can be idle or allocated: if allocated it can be travelling or working. A machine can overnight at a farm.

The frequency distribution, mean and variance of the delays are given as output; also for each machine the total work time, travel time, revenue and cost are given (on the basis of time multiplied by unit-costs).

The simulation was run with various numbers of machines, and various distributions of farms; various projection strategies were evaluated.

Subsequently, however, despite a promising start, the Bunclody initiative lapsed, due primarily to the Co-op manager falling ill at a critical period. As a consequence, owing to the uncertainty, the acreage declined and the residual farmers paid less attention to the quality of the retting process. The contract was terminated and Kirkpatricks went back, with regret, to total dependence on imported flax(10).

MILK PROJECTS

There is a link between the 'bulk milk collection' problem and the supply seasonality problem, in that the capital invested in the system is related to its peak capacity. There is also a quality problem in the off-season, with relatively small quantities of end-lactation milk handled in a system designed for ten or twenty times the volume. Contamination is a surface phenomenon and is therefore likely to be of greater importance at low volume.

Bulk Milk Collection:

Some field tests were made to get a feel for the levels of initial contamination and the extent to which the psychrotrophes were taking over after the introduction of refrigeration. These showed up a highly variable situation which, however, improved spectacularly during the course of the project (an example of the 'Heisenberg Effect'(12): word got around that something was afoot and farmers started paying attention to cleaning their vessels). Plate-counting, which had not yet become the practice at the Co-op quality control lab, was sub-contracted in from the Regional Technical College in Dundalk.

The problem of electricity-loading was at that time being approached by the ESB, who were encouraging the 'ice-bank' principle to spread the load over the whole day using a low-power heat-pump. This equipment, however, was relatively expensive and the Monaghan farmers preferred to invest in small mobile systems with relatively high power consumption over a short cooling time. When all worked together at milking-time, an unacceptably high load was placed on the extended rural electricity distribution network.

Thus the ESB was then, and still remains, concerned about the situation; electronic load management systems have subsequently been introduced for use at farm level. In some cases, standby generators have been installed, so that milking machines can be operated even if there is a power-cut. In such cases, it sometimes turns out that it pays to use the standby generator to carry the peak load, rather than to pay to have the distribution network upgraded.

This suggests the need for some in-depth analysis of the rural energy economy: for example, the possibility of using the heat of the cows' milk, pumped up to an appropriate temperature, as the main energy source for the domestic and farm hot-water supply; also the possibility of using the standby (or 'peak-chopper') generator engine as a 'combined heat and power' (CHP) source. The matching of this concept to the market needs and the technological possibilities is an opportunity on the borderline of OR and systems engineering.

The following two sections were written from the perspective of 1988, well over a decade after the MSc projects were done. RJ March 2001.

Seasonality and the Processing Industry:

The problem, however, has at last begun to be recognised in the presence of the 'superlevy' and the threat of drastic CAP reform. Public opinion will no longer tolerate mountains of unwanted butter, powdered skim or beef. There is scope for another hard look at the industry, its product mix and its approach to the market. The only possible future is to get out of bulk commodity products 'produced for intervention' and go for up-market value-added products. Lack of continuity of supply is the principal obstacle to this transition.

Some tentative steps have been taken by a few small local quality cheese enterprises and by the yogurt producers. This type of activity needs enhancement. The planning of the transition is a challenge to be faced at government level: can we create an aggressive up-market quality food industry, and get our farmers to supply it reliably?

Incidentally, most calves are born in spring, so that the beef industry faces the same seasonality problems. Any new analysis of the problem, based on the techno-economic modelling of the linked agricultural and industrial systems, must look into both beef and milk as two related sub-systems, with a common interest in the shared problem of continuity of supply.

Seasonality and the Farmer:

The current EEC policy of subsidising agriculture at the expense of the consumer with a dear food policy has two main pathological effects: (a) supply is unmatched to demand and 'mountains' are generated, and (b) intensive production is encouraged, with maximal use of pesticides, antibiotics and veterinary products, which are needed to keep animals alive under high-stress conditions. (Abuse of antibiotics is on the way to becoming a public health problem, as resistant strains of pathogens build up.)

An alternative policy is feasible, whereby instead of the consumer subsidising agriculture on a volume basis, the taxpayer subsidises the family farm by paying a social wage to the farmer, who would be contracted to manage the rural environment in the public interest. Any farming activity carried out as a component of the overall rural environment management (the latter could include care of wildlife habitats, maintenance of footpaths and ramblers' rights of way) would be subject to market prices and would gain the farmer an additional economic income over and above the social wage. Intensive practices leading to mountains of unwanted produce would be replaced by traditional biological agriculture, with good soil management and crop rotations, as outlined in a recent EEC Parliamentary Report(13).

In the context of a revised CAP, the cost of transition for a proportion of the herd to autumn-calving would be reduced in proportion as the social wage assumed a greater relative importance in the farm income. The consequences of CAP reform constitute a fruitful field for the systems modeller(14).

We return now to the last of the early 1970s MSc projects to be considered here.

THE IRISH RAIL NETWORK

". . . A timetable of train movements is first established, to cater for a given regular demand for passenger and freight services. This must be so arranged that free paths and platform facilities are available at appropriate times... a locomotive is then assigned to each train... two to the heavier trains (double heading). It is desirable that each locomotive should be assigned to a circular series of trains over a period of not more than 3 days... and that the locomotive then returns to the first train in the series. This series of journeys is known as a 'link'. Allowance must be made in the links for routine maintenance... decoupling and shunting time must also be allowed for ...

"In establishing the links ... the timetable may be adjusted slightly... provided a free path is available. The number of links is the number of locomotives required to run the system.... Setting up a complete new link system takes about three man-months... the system cannot be adapted quickly to changes in policy or demand...

"The model works as far as possible in the same way as the manual system.... The information in the daily train time-table is sorted into a list of arrivals and departures ('events')... relevant information about events is stored.... For each station of interest, a sub-list of events is set up.... 'Stations of interest' include terminals.... Stations where fuel or maintenance facilities are available, and major junctions, are known as 'centres'. A train which stops at an intermediate centre for long enough to allow fuel service or change of locomotive is split into sub-trains or 'segments'...

"A list of the stock of locomotives at each centre at midnight on Sunday, and of those still running, is fed into the model. The program then works through the master-list of events... allocating suitable locomotives to departures and relegating locomotives to stock on arrival.... Choice of locomotive is made according to specified rules.... if none is available, then various actions can be taken...

"The main piece of essential input data is the train timetable. When transformed and coded, it consists of about 1200 events; 6 items of data were needed on each event; 100 centres were considered, at any given centre up to 200 events might occur...

"Preliminary runs showed that a large saving in locomotives could be made by allowing light-engine runs within the Dublin Complex. A special queue was therefore set up.... and the event-selection routine modified to regard all Dublin events as occurring at one centre..."

TRANSITION PROBLEM

It is necessary to conclude by reminding the reader that the above systems modelling was done in high-level language Fortran or PL1 on a university mainframe computer. In most, if not all, cases it could now be done without difficulty on a personal microcomputer. In no case was the software 'engineered', that is, it was understood by the person or people who had produced it, but was not readily transportable. While to some extent the programs were'structured', they would not have been done rigorously according to the rules of 'structured programming'. Thus if any of the models were to become anything more than supports for a one-off study (for which purpose they were of course useful, indeed indispensable), they would need be taken into the clients' systems along with their producers for a period of in-house development or for a protracted period of further problem analysis using the skills developed during he project. Alternatively (and this would tend to be the modern solution), they would need to be processed by a specialist software engineering group into transportable marketable packages.

In the 1972 environment, it proved possible in some cases for the students on completion of their projects to be recruited on a temporary basis for some follow-through work. This constituted good experience and in several cases it led to long term jobs in the general area of the project (though in no case with the original client).

In the case of one project (the school model) an attempt was made subsequently to 'software engineer' it, with the aid of a computer science graduate (who had not been exposed the problem). This was not successful, as the emphasis in the software engineering was on 'data-processing efficiency', as in payroll packages, rather than on making it friendly and accessible to a long-term user who was problem- (rather than technique-) orientated. The 'culture gap' between OR and computer people was recognised as a result of this episode; it remains a problem to this day.

If the full potential of modern microcomputer technology is to be taken up, it is going to be necessary for OR people get to interact with specialist developers of reliable portable software, producing marketable problem-solving packages for market niches identified by OR consultants. Alternatively OR practitioners are going to have to become their own software engineers, making use of advanced development packages where necessary.

Notes and References

2. The 'retting' process involves the preferential degradation of the cellulose, leaving the fibre intact. This used to be done anaerobically in ponds, giving a characteristic offensive odour (the 'wet-retting' process). 'Dew retting' takes place if flax is simply left to lie in the field after pulling, to be attacked by soil bacteria under moist conditions.

3. Carroll B., Conlon, T. and Herlihy, P. A study of milk collection and quality control at Lough Egish. Statistics Department, TCD, 1973 (unpublished).

4. Howlett, N., Logan, J. and O'Brien, D. The effects of the supply pattern of milk on the dairy processing industry. Statistics Department, TCD, 1972 (unpublished).

5. Hooper, L., Lyons, D. and Murphy, M. The relationship between seasonality and profitability on Irish dairy-farms. Statistics Department, TCD, 1972 (unpublished).

6. Ní Éigeartaigh, A. and Franklin, G. A deterministic simulation of locomotive allocation in the Irish rail system. Statistics Department, TCD, 1971 (unpublished).

7. Redmond, A., Roden, J. and McQuillan, D. A computer model of a school. Statistics Department, TCD, 1973 (unpublished).

8. Hickey, J., Kinsella, I. and O'Kane, C. A preliminary approach to computer model building in horticulture. Statistics Department, TCD, 1971 (unpublished).

9. Neenan, M. and Devereux, J. Some recent researches on the growing of fibre flax. Proc. RIA, 1973.

10. In 1986, a new procedure was developed in which the flax is pulled, transported and stored centrally for retting under controlled conditions, using a bacterial culture spray. This process, which avoids the uncertainties of autumnal weather, deserves to succeed.

11. These are generic names for classes of bacteria, defined by their preferred temperature range.

12. This concept has transferred itself from quantum mechanics, where Heisenberg discovered that if you observe a phenomenon, by the very act of observing it you interfere with it; the famous 'uncertainty principle' follows. In the present context, the author is talking about a macroscopic analogue.

13. du Vivier, R. Agriculture and the environment, European Parliament, A2207185, February 1986.

This report produces evidence, mostly from France and Germany, that the switch from intensive, industrial-style monocrop production back towards more traditional agriculture practice, with scientific support from biology (primarily ecology) rather than chemistry, while producing less output per hectare would require substantially less bought-in inputs. The net earnings per farm would be improved and the earning per worker of the expanded workforce would be unchanged. Long-term conservation of the soil is possible only with biological agriculture; under intensive.practices, the soil is reduced to the status of a devitalised mechanical substrate for chemically fed crops and tends to degrade towards dust or clay. It takes 2 to 3 years to revitalise soil under a biological regime: earthworm power replaces tractor power (ie biological soils require less mechanical effort to produce a good tilth).

The du Vivier Report constitutes a time-bomb under the EEC's Common Agricultural Policy (CAP). It is, of course, being vehemently opposed by the agri-chemical interests and to date has largely been ignored by the agricultural research establishment.

14. The author is indebted to Alan Matthews and Julian MacAirt for drawing his attention to the following relevant works:
Riemsdijk, J. van. A system of direct compensation payments to farmers. European Review of Agricultural Economics. Vol. 1, No. 2, 1973, pp. 161-189.
Tarditi, S. Price policies and European integration. In Price and market policies in European agriculture, K. Thomsen and R. Warren (Eds). University of Newcastle-upon-Tyne, 1984.
Koeke, U. and Tangermann, S. Supplementing farm price policy by direct income payments. European Review of Agricultural Economics, Vol. 4, No. 1, 1977, pp. 7-31.
Castle, B., Woltjer, E. and Pisani, E. (Socialist group of the European Parliament). Reform of the Common Agricultural Policy. Document COM (86) 199 of 2315186 of the EEC Commission. This paper discusses the impact of the Socialists' proposals on the idea of paying farmers for environmental management.

Some navigational notes:

Copyright Dr Roy Johnston 1999