metrics

An important aspect of the design of the internal logistic simulation was to decide what needed to be measured, and how the results were to be presented. A set of metrics were evolved, to meet the requirements of production staff as well as the aims of the research. An additional requirement was to use a common set of metrics, such that models could be interfaced with those constructed for supply chain simulation.

The following performance metrics were used:

Lead time (the time a part spends within the focused factory)
Fill rate (level of customer service achieved)
Work-in-progress (level of inventory found within the system)
Tied-up capital (total cost of inventory)
Resilience (ability of the system to return to normal after a disruption)
Robustness (ability of the system to resist disruption)
Utilisation (level to which resources are employed)

Lead time for the focused factory is simply the time that parts spend within the manufacturing system. It is measured from the start of the first operation until the completion of the last operation. It was decided not to include the time the raw material spent in storage, since it might have arrived early from the supply chain; this is something that was not within the control of the manufacturing system.

Lead times are recorded for every part produced, the times being stored in an array during the model run, and then exported to a Microsoft Excel spreadsheet, to be shown in graph form. The graph takes the form of a stacked area diagram, showing both the overall time and where the parts may have been delayed within the factory.

Work-In-Progress (WIP) is an important measure for a manufacturing simulation, since it shows how many parts are held within the system. These may be bulky, and expensive (both, for the business unit where TECs were produced) but some are needed to maintain an adequate customer service level and to provide a good flow of workpieces to bottleneck machines where high utilisation should be a goal. The accumulation of WIP is a clear symptom of inappropriate or inadequate logistic control. If parts are spending a great deal of time queuing, they may be superfluous – although the simulation will need to be run through a broad range of scenarios to be sure of this.

Within the model, the level of WIP was sampled every 24 hours, for each manufacturing sequence and each output buffer. The results are exported to Microsoft Excel, producing graphs of the kind shown below.

Fill rate is the ability of a manufacturing system to satisfy the demands of customers. It is somewhat at odds with other goals such as reducing the level of work-in-progress in the system. It is, however, an important measure in that it makes a case for investment in safety stocks or surplus capacity when a pure ‘optimisation’ might produce a fragile system that is easily disrupted.

We measured the fill rate on a weekly basis, taking into account any backlog from previous weeks. Thus, a fill rate of 100% is desired, week after week... although real world events such as material shortages or machine breakdowns are seen to disrupt this from time to time. The graph below shows an example, where supply problems cause a temporary shortfall in deliveries.

Tied-up capital is an important indicator of the leanness of the facility. Lean manufacturing demands that excess inventory be eliminated, since this is a source of waste. It is therefore necessary to have a method of determining the cost of the WIP, while exploring the impact of reducing the levels allowed to accumulate within the system.

In the simulation, the value of tied-up capital is measured as an average figure for the whole period. Standard costs were supplied for each part, depending upon the manufacturing stage it had reached, including raw materials, and TECs undergoing welding, machining and assembly, plus finished goods. This yields a single figure for each TEC type, to be used to compare alternative systems of logistic control. An example is shown below.

Resilience is the ability of the manufacturing system to return to its planned position after a disruption. This is measured as number of weeks that are required until the backlog is back to zero after a disruption. The backlog is measured as the number of parts that are late, when compared to the delivery schedule.

Since the VAC focused factory is a complex system, and the demands placed upon it can vary, it would be naïve to seek an ‘optimal’ solution that assumes predictable demand, quality and machine reliability, etc. Instead, we accept that some variability is inevitable, and that there may be some value in having either surplus inventory or surplus capacity, such that the system can recover after a problem.

Resilience is closely related to robustness; the ability of a system to cope with a disruption, without suffering a fall in the service level. Under such conditions, there is no backlog to measure. Robustness is more difficult to perceive, since it depends upon the severity of the initial disruption, but clearly there will be some benefit in having a manufacturing system that is able to absorb variations in the up-stream supply chain, and not pass these on.

Utilisation is important, given the high value of some of the machine tools found within the focused factory, and the constraints they impose. High utilisation identifies a bottleneck, but low utilisation might indicate that the machine is being kept waiting for material, which may be a source of inefficiency. Four workcentres were identified as having acted as a bottleneck under certain circumstances, so the utilisation of each of these was recorded, on a weekly basis. The results can be seen below. Note how the utilisation falls when a supply shortage was introduced in week 17.