Because of a limited data set for some pump types, a significant dead band exists, such that when changing the C coefficient, no subsequent change in the percentage of pump failures is observed. This is a function of the small number of models available in that particular pump type. If more models for that pump class were included, should they exist in the marketplace, it would again change the equation that fits that data. The problem with trying to develop a precise data fit to some of these pump types is that the limited number of models available, along with the spread of the data, lead to a level of uncertainty, making a precise data fit unlikely. Further complicating the equation will not improve this precision and will make uniform implementation of the regulation more difficult.
The letter further points out that the Europump study was based on a population of more than 2,300 pumps, while the HI survey of U.S. pump manufacturers was based on a population of 2,124 pumps, using the same scope of equipment specified in the EU directive. The DOE framework document, however, states that 27,000 pump types were analyzed. HI questioned the validity and source of this number because it did not appear to match the proposed scope of the pump efficiency rulemaking. While the HI survey did not absolutely capture the complete marketplace, HI is confident that the market is not 13 times larger as noted in the DOE’s reference to 27,000 pump types. This conclusion is strengthened because most of the major corporations identified in the DOE framework document are represented in the HI survey.
The HI data analysis also sought to ascertain whether U.S. pump industry failure rates match those of the EU. Using EU C factors to determine the failure rates within the HI survey universe, aggregate failure rates were close to the 10-percent and 40-percent failure rates, as predicted by the equation for C values of 0.1 and 0.4, respectively. The HI survey of 17 manufacturers, however, yielded different results in some of the pump categories. The failure rate for each pump category is shown in Table 1. HI believes that this anomaly resulted because failure rates for all the categories combined in this initial analysis, which was based on averaging, would be different when each pump type was compared individually within its cohort group.
To evaluate the same pumps again, on the basis of varying only the C factors in the equation, HI was able to obtain the respective 10-percent and 40-percent failure rates for the U.S. pump data. The C factors were changed to obtain the anticipated failure rates, and the data was then consistent with 10-percent and 40-percent failure rates (see Table 2).
Based on the differences in failure rates and C factors, HI determined how well the MEI equation used in the EU standard applies to the U.S. data set. Each set was plotted and visually compared to the shape of the MEI equation’s 3-D surface plot.
A quantitative analysis of the goodness of the fit for the equation is difficult because the equation is not meant to represent a fit to the data but instead a threshold value above which x percent of the data should be. Even plotting a MEI = 0.5 with 50 percent of the data points above and 50 percent of the points falling below the surface does not necessarily represent the best fit of the equation to the data because of the non-uniform distribution of the data set.
While other equations may fit the data better, they would be complicated to modify for the multiple pump categories and the two- and four-pole operating speeds. The variation of the constant C in the MEI 3-D quadratic equation represents a good data fit and a simple way to adapt the equation for different categories and operating speeds.
The modified C values to fit the North American data represent different efficiency thresholds that the U.S. pumps would need to achieve when compared to their EU counterparts. The percent efficiency difference for each pump type is shown in the Table 3. The values in parenthesis indicate instances in which the EU efficiency was greater than the North American efficiency.
Based on HI analysis, the MEI equation is a good fit for both U.S. and EU data. The equation only requires the adjustment of the C coefficient to correctly represent the proper threshold efficiencies for the various pump types and appears to be only slightly different for the EU and North American data sets. This approach achieves global harmonization of pump efficiency standards and provides a method that appears to be both technologically feasible and economically justified.