Most MRP systems cannot handle production scheduling.It is now recognised that the majority of MRP systems cannot address the increasingly important issue of production scheduling. Usually, they produce an `infinite capacity’ plan – one which assumes that the manufacturing plant has an infinite capacity.
In reality, of course, all plants have a finite capacity, and it is the job of specialised finite capacity scheduling systems to determine the best factory loading, taking into account all the relevant constraints and business objectives.
Why do we need scheduling systems? Symptoms of poor scheduling are all too common – examples include late, unpredictable deliveries, high stock levels, low plant utilisation, reduction of effective shelf life, high production costs and reduced quality.
The problem is that the calculation of an effective schedule which can actually be used by a factory is an extremely complex, time-consuming activity – so complex in fact that a computer system is usually essential. Certainly, finite capacity scheduling is an important issue.
For example the DTI’s Advanced Control Technology Transfer (ACTT) Programme concentrates on twelve core technologies, of which scheduling is currently the one with the most people registering an interest.
Horses for courses
There are many such finite capacity scheduling systems available, with widely differing capabilities. However, it is vital that an appropriate system is used.
Process manufacture requires radically differing functionality – a system which may be applicable to a car plant will probably be useless in a chemical factory.
Discrete manufacture is typically characterised by a production line – a linear process where the object being assembled moves from one stage to the next.
Process manufacture differs from this – usually a chemical plant has many vessels, reactors, separators and other items, which can be connected in many ways to provide a large number of routes through the process. In one large brewery application the scheduling system can identify 47,000 different routes!
Clearly, not all possible routes are equally effective; for example some may require additional cleaning processes, or use reactors of an inappropriate size. But it is essential that the system can model all these routes.
Further, that model must be sufficiently detailed to enable a true representation of the plant. Inter-vessel routing is a common problem here – many scheduling systems can model the existence of plant items and the routes between them, but in a simplistic way as shown.
Each route is shown as a single line. In practice, most vessels are not connected in that way but via manifolds, as shown alongside. This is a major difference; in the left hand diagram all routes can be active simultaneously. On the right (often the real world) transfer from V1 to R1 is possible, whilst also transferring from V3 to R3, but V1 to R3 transfer would occupy the entire manifold, and simultaneous V3 to R1 transfer would be impossible.
This leads to increased demands on the modelling capability which reduces the number of possible system solutions dramatically. Scheduling Technology Group’s OPT21 and ABB’s Batch FCS are amongst those which can address this.
Another issue which affects discrete and process scheduling alike is optimisation. But, what do we actually mean by `optimisation’? We actually mean that we create the schedule so that it most nearly meets the ideal business objectives. These can change with time; a few years ago, in the depths of recession, many companies had surplus production capacity, and the survival strategy was to minimise production costs.
Others took the view that improved customer service would ensure that their market share was maintained or improved, and improved delivery performance was the most important issue.
Today, demand for most products has increased sharply, and many companies need to maximise throughput. There are clearly mutually incompatible objectives which may all, at differing times, be the current goal.
It may be that the situation is more complex than this, with the need to aim for `a bit’ of cost reduction, whilst maintaining `adequate’ delivery times and getting `quite a lot’ from the plant – a sort of fuzzy logic approach to the problem.
Research is underway in this area, with some very interesting applications of multi-objective genetic algorithms, for example, being produced at the University of Sheffield.
Other systems have complex algorithms which permit the user to change the relative importance of pre-defined key objectives. Scheduling Technology Group’s OPT21 system is an example here.
Obviously, it is imperative to select an appropriate technology for the process – one which will be of benefit to all the diverse stakeholders in the scheduling problem.
This includes production planners and managers, as well as sales, marketing, financial, maintenance and distribution issues. Suppliers and customers are also affected by scheduling efficiency.
Ensuring that the objectives and issues are identified, understood and agreed is essential.
* The Author is the DTI’s Advanced Control co-ordinator and a consultant.