CAPÍTULO 5. ENSAMBLADO, EMPAQUETADO Y TRANSPORTE DE UN
6.1 Conclusiones y discusión por capítulo
6.1.5. Conclusiones y discusión del capítulo 5
Adaptation of the IDEAS process to develop a CMWL measurement tool requires a simple extension of the approach. An index of the CMWL for a task being performed in a particular context needs to be developed as a function of the Performance Influencing Factors in that situation. In many cases, the model will involve the decomposition of the PIFs into sub-factors to a level at which they can be reliably measured. A simple example of this approach is shown in Figure 34. The following diagrams are taken from an existing computer based Influence Diagram modelling tool called IDEAS, which has
1 (best case situation, all PIFs optimal). This assessment appears in the top left of the top box of the model. The ratings of the PIFs are on a scale from 1 to 100. In the context of CMWL measurement, the SLI is interpreted in a different manner from the its original meaning as an index of the quality of the PSFs that determine the likelihood of human error. Instead, the SLI is used as a workload index, where SLI stands for ‘Seafarers Loading Index’.
There is a direct link between this interpretation and the more general meaning of the SLI discussed in previous sections. The SLI can be seen as a loading index, which is also related to the likelihood of error arising from overload. There is no intrinsic difference in this interpretation and that discussed in the medical scenario, except that the SLI is only considering factors that contribute directly to workload, rather than including other factors that may not be related to workload but which could still contribute to error probability.
Actually, nearly all factors that could affect error probability can be mapped onto a workload scale. This is because in nearly all tasks, a mismatch between demands and resources is a major error mode. This particularly the case in tasks of a dynamic nature where systems such as cars, trains and ships are being controlled dynamically by one or more operator. This generalisation is less true in tasks where underload and overload is not likely to occur, but diagnostic errors arising from a misinterpretation of a situation are possible. Given this interpretation, we can use a similar calibration approach to converting a particular workload SLI to an error probability as we have used for the more general interpretation of the SLI discussed earlier (see Figure 33).
Figure 34 Example of an ID for High CMWL situations
In the simplified example shown in Figure 34, the CMWL for a high demand task is influenced by three primary PIFs: Task demands, Personnel resources, and PIFs that increase the severity of demands These primary PIFs are decomposed into sub-factors. The numerical values in Figure 34 are interpreted as follows, using the PIFs box on the right as an example.
The quality of PIFs assessed by the derived rating in the PIF box is determined by the quality of the automation systems (e.g. radar, GPS), the weather conditions (more severe weather means that the complexity of information processing requirements in general will increase) and the level of fatigue (efficiency of cognitive processing will degrade at high fatigue levels). The ratings assigned to the sub-factors can be interpreted as follows:
Factor Initial
Ratings Interpretation Range assumptions Recalculated Ratings Distractions 80 There is a high level of distractions (1=Best case 100=worst case) 80 Automation
quality 50 The quality of the bridge automation is average (1=worst case, 100=best case) 50 Extent of Good Weather Conditions 70 The weather conditions are good (1=worst case, 100=best case) 30
Fatigue 90 The bridge
personnel are very tired (1=Best case 100=worst case) 90
Figure 35 Assessments of PIFs in Figure 34 showing scale reversals
Since the PIF box in Figure 34 refers to PIFs that increase loading, as this rating increases, the loading on the person, and hence the error rate, is assumed to increase. Figure 34 shows four factors as influencing the PIF box, Distractions, Automation quality, presence of good weather conditions and Fatigue. As some of these scales increase, they have a positive (decreasing) effect on workload. Examples in Figure 34 are Automation quality and Good weather conditions. These ratings therefore need to be subtracted from 100 in order to ensure that the derived rating (62) moves in the correct direction as ratings on this scale increase. This reversal of scales is shown column 4 of Figure 35, and is carried out automatically by the software if a minus sign is appended to the weight (see Figure 34)
The other factors such as Distractions and fatigue will increase the loading as their rating increase. These scales therefore do not have to be reversed. The derived rating for the quality of the PIFs is calculated as follows:
Rating for quality of PIFs Box = (80+50+30+90)/4 = 62.5 (rounded to 62 as shown in Figure 34)
Let us assume that the SLI scale generated by the factors in the ID in Figure 34 is related to error probability in the following manner (based on field observations):
Conditions Error rate Worst case (SLI=1) 0.1 (100 failures per 1000 demands
Best case (SLI=0) 0.001 (1 failure per 1000 demands)
Figure 36 Calibration data for Figure 34
If the SLI model is calibrated using these data, the expected failure rate based on an overall SLI of 0.56 is 56 failures per 1000 demands (see Figure 34). If the fatigue level is reduced to its best case (a rating of 1), the predicted number of failures due to overload decreases to 49 per 1000 demands. Other similar analyses are possible to investigate the changes in loading and error rates as a function of varying the assessed conditions in the model. A great strength of the IDEAS approach is that it allows the model developed by the experts to be verified by including some known scenarios in the assessment, and seeing if the predicted workload is in accordance with their experience. The ‘What if’ capability also allows cost benefit analyses to be performed if alternative load management approaches are being investigated.
5.6.7 Conclusions on the applications of IDEAS to CMWL