Tap-water expected values were provided from historical analysis reported by the municipality and by results of previous validations.
For the rest of variables, the authors referred to process equipment and components requirements, as well as the recognized
monographs for pharmaceutical water quality in the case of outlet water product.
Lab analytical equipment was used for offline measurements, such as pHmeter Orion model 30, conductimeter Crison model micro
CM2201, total organic carbon (TOC) analyzer Anatoc, LAL analyzer Biowhittaker model Kinetic-QCL, as well as chemical and microbiological
Results from fault tree analysis application
Figure 2 illustrates the main starting section of the FTA model, and Figures 3 and 4 show two follow branches. The top event
is defined as "bad performance of WPP process," from where it succeeds any malfunction event in obtaining proper quality specifications
or other deficiencies from the same process regarding its original design.
Figure 2 (FIGURE IS COURTESY OF THE AUTHORS)
To determine quality specification failures, noncompliance of water-product chemical requirements was analyzed, particularly
EC and TOC content. Noncompliance of microbiological requirements was also analyzed, specifically bioburden and endotoxin
contents. Other fault events were part of the model, including deficient softened-water handling (taking into account storage
and distribution failures), impossibility of reaching water-product delivering capacity because of scale and biofilm formation
on RO membranes, and working pressures and permeate recovery percentage deviations from their normal values caused by bad
Figure 3 (FIGURE IS COURTESY OF THE AUTHORS)
Other fault events include softeners battery malfunction and bad NaOH solution dosage that lead to high water product EC in
a short time. Free-chlorine deficit through the process is considered a limiting bioburden control, and free-chlorine removal
stage malfunctioning can cause RO membrane deterioration.
Figure 4 (FIGURE IS COURTESY OF THE AUTHORS)
Barriers also were included in the model as conditional events restricting fault-event occurrence, which in principle are
process controls. So any failure or negligence on these controls can expose the process to intermediate failures. This is
the case of:
- Checking of RO system operational pressure distribution
- Automatic checking of water product EC and its return to the water-softened tank through a three-way on–off valve in case
of an OOL detection
- Free-chlorine checking at various points of the process and specially at the RO first-stage inlet
- Checking of cleaning and disinfection performance on equipment prone to generate microbiological contamination
- Visual checking of pressure-group functioning and other automatic devices related to softened-water handling.
An evident complex dependence between intermediate events was observed, which made a model probabilistic solution difficult.
For instance, a softener's battery malfunction resulting from any failure can provoke, in a relatively short time, a potential
OOL in water-product EC because of ionic content excess, provided that there is a failure in the automatic return device.
Moreover, in the long term, this can cause RO membranes scaling in the case of corresponding control failures, then an overpressure
on membranes is generated with a consequent airtightness failure in RO modules as a result of O-ring gaskets disruption. Therefore,
streammixing can lead to OOLs in all water-product specifications, without discarding a delivering capacity decrease because
of permeate area reduction in membranes.
Historical information collected as a result of process supervision was insufficient for determining in good extent occurrence
frequencies of basic independent events. Previous information was available for some other processes, but the authors decided
not to use them in the present case because of the risk of not fitting the WPP process properly.