Fibre failures in the spotlight
European microporous membrane plant growth is rocketing. Memcor's Brian Birkenhead discusses operational issues in the production of drinking water by hollow fibre membranes
Waterborne disease outbreaks attributed to disinfectant-resistant organisms such as Cryptosporidium, Giardia, microbial spores and nematodes in drinking water have been the main impetus behind European microporous membrane (MF and UF) plant growth. From zero 15 years ago, the installed output in Europe has grown to about 2Mm3/d in 2003.
Plant operators and regulators would like a single reliable online method, such as particle counting or turbidity measurement, to provide integrity assurance.
In reality, the conventional online methods are not sufficiently sensitive to quantify integrity (generally expressed as a log reduction value – LRV) at the levels required and they will not protect an operator from risks of non-compliance. Regulators seeking to minimise the pathogen risk see membranes can and will fail and require these conditions are allowed for and detectable. The level and reliability of the system’s integrity then becomes the key issue in operation.
Membrane defects that arise during manufacture, shipment and installation, despite effective QA, are usually corrected during the commissioning of the plant.
Despite this, there will be a continuing low level of failure. The frequency of fibre failure will vary with each manufacturer’s experience and design. The impact that a single failure may have on the integrity of a production unit is a function of the system’s design and operation.
feed and Filters
The effects are to create a flow of unfiltered feed (bypass) through defects, carrying possible pathogens into the filtrate. System design factors include the fibre diameter and number of fibres per membrane module (for example, element or cartridge), the size of the group of modules which can be tested, the mode of operation and the location of the fibre break.
The greater the number of fibres required to produce a given output of water, the smaller the impact of bypass flow from a single break.
Bypass is considered arising from flows through a cylinder as in Figure 1 (Johnson et al, 2002). For small diameter fibres the formula used is based on the Hagen-Poiseuille (HP) equation used widely to describe fluid flow through pipes:
In using this equation, the worst case scenario is assumed where a fibre breaks cleanly at the pot and the break exposes the largest hole for the shortest distance.
The HP equation works well for small diameter fibres where flow results in low Reynolds numbers up to 600.
With an allowance for fibre entrance and exit losses, the HP equation can be used up to a Reynolds number of 1,200.
Above 1,200 the Darcy-Weisbach equation, incorporating entrance and exit loss, matches experimental data more closely:
Figure 2 shows the actual measured flow through single fibres of four different diameters, compared to the calculated flow. Calculations were made using the Hagen-Poiseuille and Darcy-Weisbach equations, both including an allowance for entrance and exit loss. The worst case bypass flow resulting from bypass from fibres cut cleanly at the pot can therefore be accurately predicted using these formulas, to within about 5% of the actual.
No matter which equation applies, the dominant influence on the bypass flow rate
is diameter. This is not necessarily something that can be easily changed for integrity purposes. There may be good reasons for choosing particular fibre geometry related, for example, to hydraulic distribution within a module and overall performance.
Nevertheless, fibre diameter must be recognised in the design for operation. The integrity testing of operational plants has been addressed in a number of ways, including the following:
- online instruments – many studies have now demonstrated that current online instruments, for example, turbidity meters and particle counters or detectors, are not yet sensitive enough. A USEPA report in 2001 notes particle counters, although the more sensitive of these indirect monitoring methods, have a number of well-established problems that can potentially distort both accuracy and precision of their measurements,
- challenge tests – a Bacillus spore challenge test on a 36,000m3/dMemcor pressurised CMF plant was used by the Tauranga District Council in New Zealand (Nowacki, 1998) as the basis for granting compliance. The test demonstrated the absence of spores in the filtrate but was clearly highly impractical for routine use.
Other online challenge tests involve feeding inert powder into the upstream side of a filtration unit in controlled doses for short periods.
An increase in particle count in the filtrate indicates the presence of one or more compromised fibres.
However, this type of test still relies on the acknowledged weaknesses of particle counters and there are some disadvantages related to filtrate contamination.
The smallest portion of a membrane filtration system, which can be subjected to a pressure decay test (PDT) will, for Memcor equipment, be a unit (skid) for the pressurised CMF process or a cell of submerged membranes for the CMF-S process. Each cell or unit is taken off-line for six minutes to allow the automatic test to take place. How often it is carried out may depend on the local regulator’s attitude but a frequency range of every one to seven days is usual.
After any early defects are dealt with, the fibre breakage rate for Memcor membranes settles down at a low rate of about I hollow fibre in 50M/d.
The distribution of the pattern of breakages follows the Poisson distribution for rare events (Hall et al, 2003). For a new site the estimated failure rate can be used to predict performance until sufficient site data is available.
The period between maintenance interventions for repair will be set to avoid a significant risk of the plant moving outside of acceptable integrity limits. The routine PDT would continue to be monitored for any signs, which would indicate the need for additional intervention. The sequence for this type of analysis is:
The required log reduction value is established. This is equal to the logarithm10 of production flow divided by the maximum bypass flow, so the latter flow can be calculated. The number of fibres cut at the pot, which would produce this bypass flow, is deduced. The total number of hollow fibres in operation together is calculated.
A suitable confidence level must be set. For example, a 95% level would imply a 5% probability of intervention being required before planned maintenance. The Poisson distribution in its simplest form is given by:
where P(X = a) is the probability of ‘X’ having the value ‘a’ if the expected value or mean is ‘m’. For our purposes, a cumulative version of the equation must be used to allow all numbers of breakages over the limit to be considered. An example is shown below in Table 1 for a 120,000m3/d drinking water plant. There are eight cells in this submerged Memcor CMF-S plant, each containing 396 module elements of 14,500 hollow fibres.
This can now be used to plan intervention work in a way that will minimise the need for unplanned intervention. In practice, it is likely a safety margin will be required.
Here, the maintenance could be planned at six-month intervals to maintain a LRV of 4, which allows for the fibre breakage rate to be twice as high as expected.
A 12-month interval might be possible but on a plant of this size there is an argument to divide the work into two interventions per year. If an LRV of 4.5 were needed, then an interval of three to four months could be used.
For a small installation the position is different. The failure of a fibre presents the same bypass flow but the flow of treated water forward is smaller so the event has a greater impact.
A second example uses 96 module elements in a single Memcor CMF-S submerged membrane cell to produce 3,500m3/day. The capacity to monitor integrity of an operating membrane plant is fundamental and must be addressed at the design stage.
The pressurised airflow tests (bubble-point tests) such as the pressure decay test described are the most sensitive tests and they are being adopted around the world for their simplicity and reliability.
They provide a proven measure of system log rejection. Differences in individual membrane types will affect integrity management, particularly membrane fibre internal diameter and fibre breakage rate. Statistical analysis of membrane failures can provide a useful predictive planning tool. Different scale of plant size is also a factor – small plants will need a higher intervention rate to ensure integrity is maintained.
Virus rejection is not simple to verify. There is currently no practical on-site method for testing an operating membrane plant for virus integrity.
Until a suitable test has been developed, it may be necessary to remove membranes for laboratory checks to be certain effective virus rejection has been retained