Hydraulic and related fluids are used in a wide range of industrial situations where prevention of fire is critical. Of particular concern are systems used in tunneling and mining. A number of incidents have occurred in which mineral oil-based fluids have provided a major fuel input to large fires. For example, the fire on the Kaprun funicular railway.1 The industry response has been to develop fire-resistant (FR) fluids, usually by either adding water to the oil to create an oil/water emulsion that is used as the hydraulic fluid, or by engineering chemically fire-resistant fluids. The use of fire suppression equipment, with conventional mineral hydraulic oils, is also an option that has been used and may be available in some industries.
Fire situations may arise from a variety of causes. For example, fire may result from fluid released as a high-pressure spray or as a static pool. There is no one standard test which can reflect the full range of hazard scenarios. To fully assess the fire hazards of such fluids requires a wide-ranging test program. The situation is complicated by the fact that, in the industry, there is no common understanding of the term fire-resistant and no universally accepted test(s) to measure it. As a result, fluids have been accepted as fire-resistant and selected for safety critical end-use situations according to their performance in single small-scale tests of questionable scientific basis. Typical are spray ignition and burn tests.
With the worldwide push toward goal-setting safety regulation, a new approach is required in which fluid selection is based on a risk-based approach. This has a number of advantages. It can incorporate information from a range of tests and also allow for incorporation of other methods of engineering adequate fire safety standards such as shielding of ignition sources or automatic fire-fighting systems. This places new requirements on fire testing to provide the data needed for risk assessments as well as a new approach to fluid selection.
This article explores this situation where materials have previously been prescriptively selected by reference to a standard test. It presents the findings of a comprehensive and specially adapted test program on a range of generic fluid types commonly encountered in the United Kingdom, to better quantify the fire resistance of hydraulic fluids. Further, it explores the use of these test results in a risk assessment-based fluid selection process.
Table 1. Fluid Ignition Test Program
Experimental Test Program
The range of fluids selected for the test program represents all generic types found in the United Kingdom. This included a mineral oil (labeled 123), water glycol (122), oil-water emulsion (124), phosphate ester (128), polyglycolether (127), polyol ester (126) and rapeseed oil derivative (125).
The test program was selected using the following criteria:
An attempt was made to reflect end-use conditions, particularly the types of releases which could occur on hydraulic systems including sprays at typical high pressures, fluid soaked into absorbent substrates and static pools.
The use of ignition sources of representative sizes and energies including open flames, hot surface and electrical sparks.
The tests were required to provide a range of measured data on the ease of ignition and the major fire hazards: rate of fire growth, rate of heat release and the generation rates of smoke and important toxic species.
A test program was assembled. The tests for ignition are shown in Table 1. Other tests were carried out to quantify the consequences of burning of an ignited fluid. These tests were intended to measure the rate of heat release and the smoke and toxic species production rates from burning fluid. Tests were performed at small scale in the laboratory and also at a medium scale in some of the Health and Safety Laboratory's (HSL's) large-scale experimental facilities. They comprised:
Standard test protocols, for example ISO14935, a small-scale laboratory test measuring the rate of flame propagation on a fluid-soaked wick.
Standard test methodologies extended where it was considered they could provide additional data for quantifying a particular fluid hazard. One example of this is ISO20823 - the hot manifold ignition test. In standard format this uses a single hot manifold temperature of 704°C with a pass/fail test criterion. Further information on the hot surface ignition temperature of fluids can be obtained if the manifold temperature varies over a range.
A series of ad hoc tests specially designed to reflect fluid end-use situations and provide the information required for a risk assessment. The pool fire combustion tests are an example. These were carried out in HSL's experimental fire tunnel and first examined the ignition of fluid pools ~1m2 with a range of flaming sources going on to measure the rates of heat release, smoke, CO and CO2 production.
Following this program, the tests were critically reviewed with regard to their scientific basis and the utility of the data for use in a risk assessment. On this basis, suggestions have been made for the modification of standard procedures, or new tests, to provide the information of the type required for the risk assessment process.2
Table 2. Results from the Pool Fire Combustion Test Program
Examples of Test Data
The data from one of the tests are used to illustrate the type of information provided by the tests and that required for use in risk assessment.
The test example concerns the consequences of bulk fluid combustion determined by monitoring conditions downstream of well-ventilated pool fires of 15 liters of fluid carried out in an experimental fire tunnel. Sufficient downstream measurements were taken to determine production rates of heat release, smoke and toxic species. The smoke measurements employed an array of three white light opacimeters calibrated using a set of neutral density filters to allow conversion of data to optical smoke density per meter. The temperature and flow velocity was measured at the same locations using respectively 2.5 mm diameter radiation shielded K-type thermocouples and three vortex shedding anemometers. A gas sample taken beyond the point where the flow was well-mixed, was analyzed for carbon monoxide, dioxide and oxygen.
The data were converted to engineering units and periods of steady burning identified, typically at least three minutes, to allow the computation of characteristic averaged values for heat release rate and smoke and toxic gas production. These are summarized in Table 2 and show such substantial differences between fluids to allow fluid ranking. For some fluids it must be realized that there are other, possibly more appropriate toxic products, to indicate toxic production potential.
Table 3. Relative Ranking of Fire Performance of Fluids
Use of Test Results
Ranking fluids according to the hazard they present can be performed on several different levels. One method is consideration of individual test parameters with particular reference to fluid end use. All that is required is a test which adequately discriminates between all fluids and a measurement of a parameter representative of the hazard. Of the tests carried out, most provided adequate discrimination. These include:
Ignition: hot manifold test, AIT, spark ignition and propensity to thermal runaway
Consequence: New Buxton Spray and pool fire tests.
Some other tests provided inadequate discrimination over the entire range of fluids but some limited discrimination which could be of use in a risk assessment for which data are scarce.
In this approach, the fluids are simply ranked in the order they performed in each test. Thus the best performing fluid would score 1 and the worst 7, etc. No weightings were applied to any test result. These plots can be used to provide a qualitative idea of the relative performance of each fluid over the entire test or a restricted test program.
Calculation of mean scores for each parameter/fluid provides the ranking over the entire test program as shown in Table 3. These figures were simply obtained by adding the rankings for each fluid in each test and calculating the average for the entire range of test situations. The results in Table 3 are arranged in ranking order with the most fire-resistant fluid on the left. According to this method, the best performers are the aqueous fluids followed by the phosphate ester.
Application of Risk Assessment
Three types of information are required for a generalized risk assessment:
The probability that a release may occur and the release characteristics
The likelihood that this will ignite
The consequences of fluid combustion including the resulting temperatures, smoke obscuration and toxic gas concentrations
The difficulty often encountered is the lack of data for many of the inputs. The results from laboratory tests have the potential to alleviate some of these problems, but it is not always clear how they can be used.
The probability and characteristics of a release are not fire issues, but traditionally these have been specified by reference to historical data and use of simple engineering calculations respectively. The specification of the probability of an ignition is the most difficult area. In a full risk assessment, the machine and its operational environment might be examined in detail to derive a full inventory of potential ignition sources. This would include the location and temperature of any hot surface, the presence of lagged pipes or the existence of unsealed electrical equipment, for example. Thus, in normal operation, if surface temperatures were below the hot manifold ignition temperature, then that source may be discounted in all but fault conditions. Similarly, if naked flames are not allowed in the machine operating environment, the possibility of open flame ignition might be discounted, or at least reduced to a low level. As in many risk assessment situations, expert judgment must be invoked. Thus, a probability of an ignition might be assigned by considering both test data and the machine environment together.
Consider for example, hot surface ignition. A scheme might be developed in the following form:
For a test hot manifold temperature Te. (If normal temperatures on the machine exceed this, the ignition probability is 1).
Down to normal operating temperatures below this, the ignition probability might be assigned from a sliding scale down to zero at some fraction of Te, say 0.5 or 0.75 Te.
Some additional allowance would need to be made for the possibility of a fault condition.
The consequences of fluid combustion are more easily incorporated within the framework of a risk assessment. Thus the tests provide explicit measurements of parameters such as heat release rate, smoke or CO production. The rate of smoke propagation and loss of visibility due to smoke inside a tunnel or machinery space may be computed from a knowledge of the rates of heat release and smoke evolution as illustrated in a full risk assessment on hydraulic systems carried out by Thyer.4
In many, cases a full risk assessment may not be justified. In this situation it may be appropriate to apply risk index methods, such as those described by McBride.3 In these methods, a parameter variously described as a "risk score" or "risk value", S, is computed according to the equation:
A number of attributes, for example smoke production, ignition temperature, n, are chosen to characterize each situation based on fire scenarios or loss statistics and weights assigned to them through an assessment of probabilities. The possible accident scenarios may be chosen and wj and rj aligned with likelihood of occurrence and consequences. In these schemes, numerical values are introduced for wj and rj using normalized scales. In this case, the likelihood of occurrence might be assessed using a scale from 0 to 3 with 0 being not credible, 1 unlikely, 2 medium probability and 3 highly likely. Similarly, combustion consequences could follow a similar scheme with 1 representing minimal consequences, 2 medium level effects and 3 severe consequences. The test results obtained might then be used to define these categories.
In applying this methodology, consider the example of comparing two potential fluids, a polyglycolether (PGE) and a rapeseed oil derivative in a specific-end use situation in which releases could occur as a spray and be absorbed on a substrate. The spray hazards are taken to comprise heat release and smoke production, and the ignition sources present are hot surfaces at about 400°C and sparks of unknown energy. For absorbed fluid, the only hazard is the ability to propagate a flame. For hot surfaces, the rapeseed will probably not ignite because its hot manifold temperature is ~466°C. A conservative approach might include the possibility of ignition, so a weighting factor of 1 may be appropriate. For PGE, the value was 396°C and the ignition probability taken as one with a weight factor of 3. The spark ignition energy for the rapeseed was found to be low (<1J) whereas that for the PGE was an order of magnitude higher. As a result, the rapeseed and PGE should be graded 3 and 2, respectively, for this aspect. Because the tests on thermal runaway suggest that both fluids would respond at such temperatures (onset of thermal runaway at 190° and 170° for the rapeseed and PGE, respectively) a weight factor of 3 should be assigned to each fluid. For spray combustion consequences, the gradings for heat release are 3 for the rapeseed and 2 for the PGE and for smoke production 2 and 3. Both fluids supported flame spread on a wick with similar rates; they should perhaps both score 3 on this aspect. The matrix, Table 4, provides a method of visualizing how the overall risk rating may be assigned using the PGE as an example. Summing the scores suggests rapeseed has slight advantage over PGE with totals of 29 and 34 for the rapeseed and PGE, respectively.
Table 4. Matrix of Risk Ranking for PGE Fluid
The study has had some success in identifying a number of tests which could be applied across the full fluid range and showed good discrimination between all fluids. Some tests provide data suitable for direct incorporation in a risk assessment, but these often require considerable manipulation and interpretation because the test conditions rarely correspond to the end-use environment and expert judgment must be applied. Care must be taken to ensure that this judgment produces a conservative result.
The simplest way to use the test data is by fluid ranking on the basis of a single test parameter. To combine fluid performance over a range of tests, and therefore release and ignition scenarios, is a more difficult process. This is best carried out within the framework of a risk assessment. Most rigorously this might be achieved through a full quantitative risk assessment. In many cases, this may not be appropriate. A simple semiquantitative risk assessment process has been developed giving a basis for the development of a fluid classification scheme. It involves the interpretation of the test results for probability of ignition and consequences of burning the fluids in terms of performance bands. A classification scheme has been tentatively suggested as a basis for discussion. This represents a significant improvement on the current situation in which limited test data are generally employed to inform fluid selection in an unstructured way.