Three studies were conducted to determine the utility of the EAMP. First an experiment was conducted in an engine laboratory to verify the detection of an engine fault via exhaust emissions. Then the EAMP's loading method was tested at a mine equipment manufacturer's facility. Finally, the EAMP was tested in an underground mine. The emission results for all three studies are reported on a dry exhaust basis because on-site measurements are taken on a dry basis.

Laboratory evaluation

Methods

A Caterpillar 3304 PCNA engine, fueled with low-sulfur diesel, was used to evaluate the change in engine exhaust composition caused by a clogged intake air filter. The clogged filter was simulated by a butterfly valve fitted in-line with the engine's air intake. As the valve was closed, it caused a reduction in air available for combustion. Table 1 indicates the test specifications, and Figure 6 illustrates the laboratory apparatus.

Figure 6: Laboratory apparatus

DPM, CO, NOx, and CO2 emissions were monitored as engine speed, load, and air restriction were varied. DPM was calculated from opacity measured with a 4-in Celesco model 107 smoke meter.(20) CO and CO2 were measured with a National Institute of Standards and Technology (NIST) traceable Pierburg BINOS 2000 nondispersive infrared analyzer (NDIR), and NOx was measured with a NIST traceable Pierburg PM 2000 chemiluminescence detector (CLD). The test matrix consisted of "clean," "dirty," and "clogged" air-filtering conditions. At each of the three filtering conditions, the engine load was varied between 90% and 100%, and the engine was operated at both 1,600 and 2,100 rpm (Table 2).

Results

Ideally, on-site emission testing would directly measure DPM as a fault-indicating emission and equivalence ratio (f) as a load-indicating parameter. DPM indicates many common engine faults, and it is a significant emission from a health and safety perspective. f is the actual fuel-air ratio divided by the stoichiometric fuel-air ratio, and it is proportional to engine load. f is a more useful parameter than fuel-air ratio because it is independent of fuel chemistry, and a f of 1 always corresponds to zero excess air.  Figure 7 illustrates how nonvolatile DPM (DPMnvol) responds to changes in f at different speeds and intake restrictions, and how it can be used to determine if an engine requires maintenance. The "pass," "warn," and "fail" lines in Figure 7 (and Figures 10, 16, and 17) are suggested guidelines for the model 3304 PCNA engine. The lines were arbitrarily set for this experiment and should not be interpreted as any requirement. The "clean", "dirty", and "clogged" arrows in Figure 7 (and in Figures 10, 16, and 17) indicate the maximum load achieved at each filtering condition.

Unfortunately, DPM and f are both difficult to measure outside of a laboratory. However, CO and CO2 can be used as surrogate indicators of DPM and f respectively. Figure 8 illustrates the linear relationship between CO and DPM at high loads, for the 3304 PCNA engine. It was not determined how this relationship may vary from engine model to engine model. A more complete discussion of the statistics in Figures 8 and 9 can be found in the "Statistical Analysis of Data" section of this report.

Figure 8: Relationship between CO and nonvolatile DPM at high load

Figure 9 demonstrates the linear relationship between the CO2 data and f. The relationship between CO2 and f can be calculated as well, if the H:Cratio of the fuel is known. The stoichiometric (f = 1) complete combustion equation for a hydrocarbon fuel in air is

and the stoichiometric dry concentration of exhaust CO2 is the ratio of CO2 to the dry products of combustion.

1* is the moles of CO2 formed as a product of combustion. If the amount of fuel is decreased while the amount of air is held constant, the number of moles of CO2 will decrease, but the moles of excess O2 will exactly offset the decrease in CO2. If the amount of fuel is decreased, 1* in the denominator will remain constant while 1* in the numerator will become the fraction of actual fuel to the stoichiometric amount of fuel for the air amount of air in the equation. This fraction is f.

The fuel used in this experiment had a H:Cratio of 1.73 ±2%, therefore,

This equation is plotted in Figure 9 along with the actual CO2 data. The error bars for the calculated CO2 indicate the propagated uncertainty of the H:Cratio. Figure 9 indicate that, because of the uncertainty in the H:Cratio, the calculated CO2 values agree with the actual data just as well as the best-fit straight line through the data.

Figure 10 demonstrates how CO and CO2 can be used to determine if an engine requires maintenance. High concentrations of CO indicate that an engine is probably emitting high concentrations of DPM, and high concentrations of CO2; indicate a sufficiently high equivalence ratio for a valid engine test. In practice, any single measurement taken using the EAMP could be plotted as a single point on a CO versus CO2 chart. The pass, warn, and fail lines in Figure 10 indicate that for any load, CO exhaust concentrations should not exceed a predetermined value, and for a valid EAMP test, the CO2 concentration must be above a minimum predetermined value. A fault also may be indicated if CO2 concentrations exceed a maximum value.

Portable instrument evaluation

To use CO and CO2 measurements in the field, portable emission analyzers should be accurate enough to correctly detect the clean, dirty, and clogged air-filtering conditions established in the clogged air filter experiment.  Measurements taken with two portable analyzers were compared to measurements simultaneously taken with the NIST traceable NDIR and CLD instruments.

The Ecom-AC portable exhaust gas analyzer uses electrochemical gas sensors to measure O2, CO, NO, and NO2. It calculates CO2; from O2 measurements and a selected fuel composition (H:Cratio). It also measures exhaust temperature and back-pressure. This instrument features a PC interface and a sophisticated gas conditioning system. The analyzer is housed in a rugged aluminum case (19 by 16 by 6.5 in with a total weight of 28 lb), that makes it suitable for in-mine operation. The Ecom-AC is illustrated in Figure 11.

The Ecom-E exhaust gas analyzer is a 5-lb hand-held analyzer that does not have a PC interface. It measures only CO and O2, and calculates CO2 by the same method as the Ecom-AC. The Ecom-E is shown in Figure 12.

The gas sensor specifications for both the Ecom-AC and Ecom-E are listed in Table 3.

These instruments were evaluated because of recommendations made in a 1995 U.S. Bureau of Mines report.(8)

The CO and CO2 measurements from both analyzers were compared to measurements made with the NDIR analyzer. The Ecom-AC's NO and NO2; measurements were summed and compared to measurements taken with the CLD NOxanalyzer. Tables 4 and 5 summarize the results at the maximum load data points for each filter condition. Appendix B contains all data generated during laboratory evaluation of the EAMP.

Statistical analysis of data

The reduced chi-squared (c²_red) test was used to determine whether the Ecom-AC and Ecom-E results agreed with the NDIR and CLD results. This test was also used to compare the NDIR analyzer CO2 data to the CO2 values calculated based on f and the H:Cratio of the fuel. Because all the data was drawn from the same parent distribution, namely the concentration of emissions in the engine's exhaust, the following equation was used to determine the probability that any two sets of data were gathered from the same parent distribution.(9)

where

n is the number of bins, or modes into which the data were distributed.

g_j is the mean value of the evaluated analyzer data at each mode.

h_j is the mean value of the laboratory analyzer data (or calculated value) at each mode.

s_(gj) is the uncertainty of the evaluated analyzer data at each mode.

s_(hj) is the uncertainty of the laboratory analyzer data (or calculated value) at each mode.

v is the number of degrees of freedom, which is the number of bins into which the data were distributed
minus the number of parameters that were calculated from the data to describe hj. When comparing two
separate data sets, hj is the data itself, therefore, no parameters were calculated, and v = n.

The method of least squares was used to determine the linear and exponential curve fits to the data. Goodness of fit was determined by the following equation:

where gj is the measured value and hj is the value of the curve fit. n = n –2 for the linear and exponential curve fits because each fit contains two parameters determined from the data itself.

Uncertainties in the analyzer measurements were estimated by applying the analyzer manufacturers' stated accuracy to each mean value at each mode.

A c²_red value on the order of 1 indicates that the expected values agree well with the actual values. To quantify these results, the probability of obtaining a c²_red value as large as the value actually obtained was determined from the c2 probability distribution.(9) If the probability, Pc is less than 5%, then the level of disagreement between expected and actual values is significant. Table 6 shows that none of the results indicated a significant disagreement between the portable analyzer data and the laboratory analyzer data.

Figures 13, 14, and 15 also illustrate that the portable analyzer data agrees well with the laboratory analyzer data. The linear regressions in Figures 13,14, and 15 indicate the best fit to the data.

Figures 16 and 17 illustrate the results of the clogged filter experiment, as measured by the Ecom-AC and Ecom-E analyzers. Both portable analyzers correctly identified the clean, dirty, and clogged filter conditions. A post-calibration factor was applied to the Ecom-E CO2 measurements because it was discovered that its O2 sensor was improperly zeroed.

The exponential curve fits in Figures 7, 10, 16, and 17 are evaluated in table 7. Table 7 shows that an exponential relationship between DPMnvol and f probably exists at 2,100 rpm, but probably not at 1,600 rpm, and that an exponential relationship exists between CO and CO2 at both speeds. Combustion in diesel engines is a complex process where simple exponential relationships do not fully describe the relationship between exhaust constituents. However, the exponential relationship between CO and CO2 at high loads, for the 3304 PCNA engine, indicates that expected emission concentrations can be determined, which, in turn, can be used to establish reasonable pass, warn, and fail criteria. For the Ecom-E, because only three data points per mode were recorded and because of the small number of degrees of freedom, insufficient data were available to determine c²_red values.

Conclusions

The laboratory-grade instruments identified simulated clean, dirty, and clogged air filter conditions at loads greater than 95% of full load at two different speeds. By the c²_red test, two portable analyzers agreed well with the laboratory-grade instruments, but only after proper instrument calibration. Pass, warn, and fail criteria were established for a Caterpillar 3304 PCNA engine by comparing CO and CO2 emissions. An exponential relationship was found to exist between CO and CO2 at high loads for the Caterpillar 3304 PCNA engine. More engines should be tested to determine if their CO and CO2 emission have a similar relationship.

Field evaluation of the EAMP at an equipment manufacturers facility

There were two objectives of this field study: to evaluate the torque converter stall loading method, and to test the Ecom-AC's performance in the field.

Methods

Six pieces of new diesel-powered mining equipment were tested using the EAMP at Atlas-Copco Wagner's mine equipment manufacturing facility in Portland, OR. Full load conditions were achieved by using the torque converter stall loading method. In addition, one piece of diesel mining equipment equipped with a hydrostatic transmission was partially loaded and tested. The Ecom-AC portable gas analyzer was used to measure emissions, and a laptop computer was used to display the results and log the analyzer measurements at 1-sec intervals. All six pieces of equipment were tested three times within about two hours.

A Detroit Diesel electronic control (DDEC) reader provided by Wagner was used to monitor the Detroit Diesel Corporation (DDC) engines' electronic controls. Engine speed and load (torque) were recorded from the electronic control. For the series 50 engine the recorder indicated 1,740 rpm and 733 ft-lb of torque during the torque converter stall. A hand-held photo-tachometer was used to verify engine speed, and this speed was plotted on the series 50's published engine performance chart, which indicated that the engine was operating along its designed lug curve, and that the engine control module was reading the torque correctly. Figure 18 illustrates the DDEC reader, Figure 19 illustrates the use of the photo-tachometer, and Figure 20 illustrates the series 50's performance chart and the operating point on the lug curve at which the engine was tested. Similar readings verified that the series 60 engine was also loaded along its lug curve.

For the Deutz engines, no electronic means were available to monitor engine torque directly, so the
photo-tachometer was used to verify that engine speeds were below the rated power speeds. It was assumed that if the engine speed remained below rated power speed while the fuel pedal was fully depressed, then the engine was operating along its lug curve. Table 8 shows the engine speed recorded as the EAMP was performed on each engine. Even without measuring speed with a tachometer, a technician performing the EAMP should be able to observe a speed reduction by the sound of the engine alone. This change in sound became obvious after testing the 6 engines.

Only one of the six engines was not loaded along its lug curve. The Deutz F4L912W was installed on a HST-1A Scooptram, which was equipped with a hydrostatic transmission. Because of the equipment design, it was not possible to load the engine while the equipment was stationary. An attempt was made to load the engine by operating its scoop hydraulics against the ground. This partially loaded the engine, but the engine still operated at 160 rpm above its rated power speed. It was concluded that the fuel pump's speed governor was still limiting fuel delivery, which is not desirable because detecting over-fueling and air restriction faults may be impossible unless an engine is fully loaded. Additional research will be necessary to develop a suitable method for loading some equipment equipped with hydrostatic transmissions.

Results

The results in table 9 show that different engines, even when new, have different normal, or baseline, emissions.  Because engine manufacturers set maximum fuel delivery rates to meet power and emission requirements, normal maximum CO2 concentrations vary from engine to engine. Limiting fuel delivery may also prevent the formation of excessive emissions otherwise formed as a result of an engine fault. All turbocharged engines tested had lower maximum CO2 concentrations compared to naturally aspirated engines.

Because turbochargers are driven by engine exhaust, there is a delay in their response to a change in fuel delivery rate. This is called turbocharger lag, and it affects transient emissions. When maintenance personnel perform the EAMP, they must be aware of this effect so that they do not mistake normal transient emissions for an engine fault.  Figure 21 illustrates the effect of turbocharger lag on the series 60 emissions. Only data obtained toward the end of the test may be used to evaluate engine condition. Therefore, turbocharged engines, determining the steady-state CO or NO emissions may require a close examination of the logged data.

For naturally aspirated engines, determining steady state CO and CO2 data at full load is more straightforward than for turbocharged engines. Figures 23 and 24 illustrate the rapid stabilization of CO and CO2 emissions from a naturally aspirated engine. Directly reading valid CO data from a portable analyzer's display is easier with a naturally aspirated engine than with a turbocharged engine.

Conclusions

Since the standard deviation of CO2 averaged less than 5% over all the repeated tests, it was concluded that the torque converter stall method of engine loading provided a way to load an engine to a repeatable speed at full load.  The Ecom-AC performed well in the field. It was sufficiently portable, and the emission values it measured were similar to expected values.

This field study confirmed that different engine models produce different normal emission concentrations, therefore, an engine-model-specific database of baseline emissions should be established so that EAMP results can be compared to a common set of data. This database should consist of normal CO and CO2 exhaust concentrations for each engine model subject to the MSHA emission test requirement.

Field test in an underground mine

The objectives of this field test were to evaluate engine loading and exhaust sampling methods in a mine and to determine if the portable analyzer was compatible with the mine environment. The feasibility of performing the EAMP on a weekly basis was determined. Mine personnel were trained to perform the EAMP during this field study. Engine emissions were compared to emissions from similar engines. Also, useful comparisons were made between left and right exhaust bank emissions of two V-block engines.

Methods

Ten pieces of machinery equipped with torque converters and three pieces of machinery equipped with hydrostatic transmissions were tested at an underground mine in Sudbury, Ontario. Although these tests were conducted at a metal mine, the equipment included scoop-trams, boom trucks, forklifts, personnel carriers, mine trucks, and locomotives, all of which are similar to equipment used in underground coal mines. The Ecom-AC four-gas portable analyzer was used to measure exhaust emissions. A laptop PC was connected to the Ecom-AC to log the analyzer measurements at 1-sec intervals. Figure 22 illustrates how the gas analyzer and laptop PC were set up in the field.

The loading and sampling methods described in the "Methods" section of this report were used to evaluate each piece of equipment. The steady-state, full-load condition was defined as the time interval over which the recorded CO2 emission attained its highest concentration and remained relatively constant for a period of seconds. In most cases, it was not difficult to locate this interval. The other emissions (CO, NO, NO2) were averaged over this interval. Similar CO2 values between tests on a specific engine model ensured that the test was repeated properly.

The emission results are reported in parts per million, and they are, therefore, independent of engine horsepower. A more powerful engine will emit proportionally higher mass quantities of emissions, even if exhaust concentrations are identical to exhaust concentrations of a smaller engine.

The EAMP was performed in about 5 to 7 min per engine, provided that the equipment operator warmed up the engine while the instrument operator started the analyzer, and that there was no problem locating a suitable sampling location in the exhaust system. The actual time over which data were recorded was 1-min per engine: 15-sec before the torque converter stall, 30-sec at stall, and 15-sec after stall. These time intervals ensured that emissions were completely drawn through the sampling line and analyzed by the instrument. During one of the test days, the instrument was set-up near a ramp where several pieces of production equipment were passing. Because the production equipment engines were already warmed up and the instrument was already operating, each test was completed in about 3-min per engine.

Results

Results from the torque converter equipped machinery tests are shown in table 10, and table 11 summarizes similar data for machinery equipped with hydrostatic transmissions. Comments in tables 10 and 11 must be validated by further engine diagnostic tests before steps are taken to correct any deficiencies. Most of the diesels produced expected emission concentrations, but some tests indicated that engine faults were probably present.

A mine truck was tested on different days (Table 12), and the day-to-day repeatability of the loading method was excellent. The standard deviation of the CO2 averaged less  than 5% over all the tests.

Emissions upstream and downstream of a catalyst

The emissions of the Minecat 097 were tested both upstream and downstream of the exhaust oxidation catalyst (Table 10). Because the catalyst oxidized 86% of the CO in the exhaust, it can mask faults that are indicated by high CO. Although the Ecom-AC's NO2 sensor was not calibrated before use, it does indicate a relative increase across the catalyst. This supports previous research that indicated that some catalysts oxidize NO to NO2.(14)

Loading engines with hydrostatic transmissions

The method described in the "Methods" section was used to load engines with hydrostatic transmissions. In some cases it was difficult for the equipment operator to control engine load. Figure 22 illustrates the problems in loading the Kubota forklift. The emission levels show that the engine was overloaded, stalled, and restarted when the EAMP was attempted. The results are very erratic compared to the results shown in Figures 23 and 24, which depict torque converter loading results. With practice, the equipment operator was able to control the forklift so that valid steady-state emissions could be sampled for 5-sec. However, the operator was able to control the hydrostatic transmissions of the two locomotives for up to 10-sec (Figures 25 and 26). It was not determined whether this method of engine loading was detrimental to the equipment.

Bank-to-bank emission comparisons

The EAMP was used to compare emissions between the cylinder banks of two V-block engines. The Jarvis Clark 595 mine truck was equipped with a 372 hp V-block Deutz BF12L-413FW. The first EAMP on this equipment indicated that the CO from the right bank exhaust was almost twice that from the left bank, so the test was repeated the next day to verify the result. A third repeat was performed on a third day after new air cleaners were installed on both banks. All of the tests returned similar results (Table 12). It was possible that the much longer exhaust and intake piping, installed on the right bank, restricted air flow to these cylinders and resulted in higher CO concentrations than the left bank.

A comparison of cylinder bank emissions was also made on the 674 Wagner Scooptram, equipped with a Deutz F6L413FW, V-block engine. The left bank emitted over twice the CO and 25% more CO2 than the right bank (Figures 23 and 24). This high CO and CO2 combination may indicate a dirty air cleaner on the left bank air intake. Because the right bank CO2 was only 5.1%, another cause of the discrepancy might be that some of the right bank cylinders were not receiving the same amount of fuel. The fuel pump and injectors should be tested to ensure that the fuel delivery to all cylinders is balanced.

Engine-to-engine emission comparisons

Since the two locomotives tested were both powered by Deutz F4L912W engines, their emissions were compared directly. Comparing the difference in emissions illustrates how two similar engines can be compared to identify a potential engine fault. Locomotive 376 had twice the NO and 80% of the CO as the 383 locomotive (Figures 25 and 26). This could be an indication of an over-advanced fuel injection timing fault on the 376. This potential fault probably would not have been detected if only CO and CO2 were measured.

Normal emissions from one engine may not be the normal emissions from another engine. Because naturally aspirated, direct-injection engines typically have high peak combustion temperatures, they produce high levels of NO in comparison to indirect injection or turbocharged, direct injection engines. Figure 27 compares the NO emissions from different engine configurations. The normal NO concentration from a naturally aspirated, direct injection engine might be considered excessive, if it were emitted from an indirect injection or turbocharged direct injection engine.

Modern engine designs often result in significantly lower emission concentrations than older designs. The Wagner Scooptram 388 was equipped with a DDC series 60 engine. This modern engine is known to produce low levels of CO. In comparison, the Wagner Scooptram 647 was equipped with a Caterpillar 3304 PCT engine. The 3304 PCT is an old engine design, and it is known to produce high levels of CO. In these tests, it produced ten times the CO as the series 60 (Figure 28). Design characteristics of the new series 60 engine that improve emissions include high-pressure, direct-fuel injection, high turbo-boosting, air-to-air after cooling, and electronic fuel control.

Conclusions

The EAMP can be performed easily in an underground mine, and maintenance personnel felt that the procedure was straightforward and easy to perform. The engine loading and emission sampling methods provided useful and repeatable results in the field. The Ecom-AC sustained the rigors of the mine environment, but the gas analyzer and computer combination proved to be bulky, and maintenance personnel suggested that a hand-held, data logging analyzer be evaluated in the future. Overall, the field test indicated that the EAMP can be used successfully in an underground mine.