Comparison of Three Preventive Maintenance Strategies for Steam Trap Systems

(Presented to the International District Energy Association, Portland, Oregon,
June 1994, Updated 10/97)

Frederic A. Hooper, Jr.
Ronald D. Gillette

Many steam traps are found in utility rooms, basements, crawl spaces and other uncomfortable places. When they leak, the steam losses are hidden C the leak is internal to the trap into the condensate line, which hides the steam loss. Testing them is expen-sive, but not-testing them permits deteriora-tion of the steam traps, with more energy required to provide necessary heat. This paper examines the effect of normal trap testing and replacement compared to two potential improved strategies.


Normal trap testing and replacement is to test traps each year and replace those that have failed. This strategy is easy to implement and manage. However, as the average failed trap leaks for six months before replacement, considerable steam is lost. Thus, the annual test & replace strategy is subject to two major costs: testing and steam loss. The other strategies tested are means of controlling these costs.

If annual test & replace results in six months of leakage before replacement, it appears appropriate to test traps that can leak large amounts of steam more frequently than those that can only leak small amounts of steam. In other words, if the traps are tested too frequently the testing will cost more than the steam that could be lost; if they are not tested often enough the steam loss will be greater than the cost of testing. In practice, the optimum testing interval equates the cost of testing to the expected steam loss. (The derivation of this interval is in Appendix B.) This interval will be different for traps that are of different types, sizes, pressure and condensate loads. Trap management with all traps tested at these intervals is termed "optimum test & replace".

As testing steam traps is expensive ($15-25 per test), it would be convenient if they could be replaced without testing, just before failure. Unfortunately, there is so much variation between steam systems (appropriateness of trap selection and water additives, cleanliness, pressure variation, etc.) that trap life cannot be predicted across systems. Even if there were not such variations between systems, individual trap application varies so much that comparison is difficult C for instance, if a steam trap is downstream from a motorized valve its life will be greatly extended because it isn't being used when the valve is closed. Appendix A presents variables that allow comparison of trap life across such variations in application.

Using these variables, if the average life of a particular type and size of trap is known, the life of other similar traps can be predicted. If the relative lives of traps are known, this information can be used to predict trap lives for a different steam system if the average life of a single trap in that system is known. This informa-tion is assumed to exist for the modeling of predictive maintenance. To obtain these average trap lives, samples of 20 traps are used. If the only traps tested are those in the samples and those trap types with less than 20 traps (orphans) and this information is used to replace all traps, this management strategy is termed "predictive replacement".


The test sample is 2061 steam traps located at a training and education facility on the northeastern United States, 93% are used for seasonal heating. Steam pressures are 4-99 psi with 83% below 13 psi. Impor-tant as-sumptions were $6.00 per 1000 lb. steam cost, $20 testing cost, 4-year trap life, and 20% leakage before the leak is detected. These are discussed in Appen-dix A. Detailed in Table 1 are other assumptions.

Table 1: Trap Type Frequency and Condensate Load

Number of Traps

Cond Load

No of Samples

No of Orphans

Prop Tested

 of IM

 Float & Thermostatic







 F & T High-Capacity







 Inverted Bucket







 Thermostatic Inline







 Thermostatic Angle














 Thermodynamic Low Cap







 Overall Sample







The condensate load differs by type of trap because of the application of the traps in the steam system. Samples and orphans are discussed above. "Prop Tested" is the proportion of sample traps tested at optimum intervals for predictive replacement. The "Prob of IM" is the likelihood of infant mortality. This can be different for each type of trap. Replaced traps are tested for infant mortality except for "predictive replacement", these are tested only if the potential loss during the first year is greater than the cost of testing the trap as they are not tested for the first year after replacement.

Based on the use of the steam traps in the system with the variables shown in Appendix A, the potential steam losses were calculated as shown in Appendix C. Deterministic simulation with assumed trap lives was used to model the trap and energy costs. This was done for each trap type, size and pressure combination.

 probability x (replacement + steam + IM test)
  of failure ( cost loss cost )

+ ( 1 - probability) x (testing) = annual maintenance cost
  ( of failure ) ( cost )

As the trap life was assumed to be 4 years, if the trap was in continuous use its probability of failure in any one year was 25%. Traps in less intensive use (e.g. all of the seasonal-use traps) would have a longer life and a lower annual probability of failure. The results shown below change little in relative maintenance costs if the trap life is assumed to be different.

Replacement costs were calculated using trap replacement costs derived from a competitively-bid contract. As the contract was very large, the trap replacement costs were increased 50% to represent likely costs for replacing a dozen traps at one time. For example, a 1/2" Thermostatic trap (the most common in our sample) was assumed to cost $80, including both labor and materials.


The results of the simulation are shown in table 2. In addition to total annual costs, the annual costs for the average steam trap in the sample are shown for testing, replacement, and energy loss.

Table 2: Typical Annual Steam Trap Costs

 Management Strategy





 Annual Test & Replace





 Optimum Test & Replace





 Predictive Replacement





Testing the steam traps at optimum intervals reduces the annual cost 19% below annual test & replace. Predictive replacement reduces annual costs 56% below annual test & replace. It accomplishes this by reducing both testing and energy loss costs to their economic minimums. This testing cost is the cost of testing the samples and orphans, plus any worthwhile testing for infant mortality. The minimized energy loss results from infant mortality not tested for.

Because the trap types differ both in replacement cost and potential steam losses, the annual costs and potential savings differ by trap type. As shown in Table 3, savings from optimum testing intervals range from 9.5% for thermodynamic traps to 30% for high capacity float and thermostatic traps. Savings from predictive replacement ranges from 34% for thermodynamic traps to 66% for thermostatic traps. Clearly, the potential savings for a steam trap system are affected by the types of traps used, but the savings are still high.

Table 3: Annual Maintenance Costs by Trap Type

 Trap Type

 T & R

 T & R

Predict Replace

Optimum Replace

Predict Replace

Float & Thermostatic

$ 64.60

$ 56.17

$ 39.39



F & T High-Capacity






Inverted Bucket






Thermostatic Inline






Thermostatic Angle












Thermodynamic Low Cap






Sample Average Cost

$ 44.80

$ 36.19

$ 19.59




Any consistently-applied steam trap maintenance management strategy is likely to be a great improve-ment over no maintenance strategy. However, as steam traps differ greatly in their conditions of use, it is desirable to take these differences into account when manag-ing them. In general, the more is known about the differences, the better the traps can be managed. Properly using this information is what makes such large cost reductions possible.


Appendix A: Variables in the Simulation

A basic concept of this analysis is the concept of constant use. By this, we mean that a trap has steam delivered to it every minute of every day. Two of our variables measure the deviations from this.

DUTY can be Continuous, Seasonal and Intermittent. Most of the traps in the sample are used seasonally. For the test facility, this is about 54% of the year. Thus a trap that would last 4 years in continuous use would last 7.4 years in seasonal use.

%USE is the time that steam is delivered to the trap when it is "on duty". For example, if a trap is located after a motorized valve, it may have only 40% use. A trap in seasonal DUTY and 40% use (termed "S-40") would last 18.5 years if its expected life under constant use (C-100) were 4 years.

APPlication type - Drip, Unit Heater, Fin Tube, Process Steam, Heating & Ventilation, and Tracer Line. Traps will last different lengths of time in these different applications. This distinction is not used in the present research but is necessary if its results are to be properly applied.

The average condensate load in the steam system affects the possible ener-gy losses. As the trap orifice through which the steam escapes is partial-ly blocked by condensate flow, steam cannot escape through that part. In all steam systems there will be condensate flow. While the highest losses can be est-imated by setting this to zero, these losses are not possible. Regardless of the num-ber of leaks, there will be some heat delivered and, therefore, some conden-sate flow.

The $6.00 cost of steam per 1000 lb. is a conservative estimate of the cost of steam actually delivered to steam traps. It includes energy cost plus chemicals, labor, and other costs necessary to produce and deliver the steam to the trap. As there is always some line loss, the cost of delivered steam will be greater than the cost of generating it. The amount of line loss is dependent on the length of the line, its insulation, and losses through traps on the main steam line.

The trap life of four (C-100) years is arbitrary, but within the likely range. We do not know the expected life of a trap under constant use, so any value is arbi-trary. Most likely, it differs with trap type and other cond-itions of use. Also, trap life depends on the trap being properly sized for its ap-plication. Too-large or too-small traps will begin to leak before a properly-sized trap.

The testing cost is based on prevailing wage costs and time to test a trap using infra-red or ultra-sonic testers. In many trap locations OSHA regulations require that two people be present to test the trap.


Appendix B: Optimum Checking Interval

The calculation of the optimum checking interval minimizes two opposing costs: The costs of testing and not-testing traps. The cost of testing is obvious. It is the out-of-pocket cost of testing the average trap C primarily the cost of the people involved. As this is dependent on time, if many traps are located so as to require two people to test them, testing cost will be increased.

The cost of not-testing is not obvious because steam loss is generally not obvious. The simplest way of determining this cost is to assume that the trap becomes bad halfway through the testing interval. Thus it will have an average leakage-rate of half the assumed leakage-rate to use in calculating the cost of steam lost. (See Appendix C for steam loss calculations.) However, not all traps will become bad during this time. The expected loss is the probability that the trap will become bad times the loss if it does become bad.

We find that

(Lt/2) = $ loss at leakage-rate for 1/2 the testing time and

(pt) = the probability of trap failure during testing time.


L is leakage-rate steam loss for one year,

t is the testing time in years, and

p is the probability of trap failure during one year.

Thus, we obtain the formula

C = (Lt/2) x (pt)

or, t = (2C/Lp)1/2 where C is the cost of testing the trap.

The leakage-rate should be set to 100% because of the rapid deterioration of steam traps once steam leakage begins.

Because the optimum testing interval formula assumes that all traps are C-100, testing intervals must be adjusted for lower DUTY and %USE. We divide the optimum test interval by their product. For instance, for a Seasonal DUTY trap that is in %USE 40% of the time, we would divide the "t" by 28%, thus increasing it by 257%

This optimum testing interval calculation is accurate for short testing periods C which will be found primarily for C-100 traps. Complication enters because the costs in the above formula occur at different times. The testing cost is paid at the end of the testing period, the loss is paid (on average) 3/4 of the way through the testing period. As traps with seasonal or intermittent DUTY and lower %USE should be tested less often, the cost of capital should be taken into account.

The cost of capital is the amount that could be earned by investing the money in a project of similar risk. As this investment is almost riskless, this low risk should be represented in the rate chosen. The result of the adjustment is:

t' = t x (1 + r)-t/8

where "t" is measured in years and "r" is the cost of capital (e.g. .05).

This will shorten the testing time by a small amount. While this amount is frequently inconsequential at current interest rates, it would not have been so in the early 1980s. Even today, it will be meaningful for longer testing intervals.


Appendix C: Cost of Steam Lost

Cost of steam lost is calculated in a normal manner from a variant of the Napier formula:

Cost of Steam Lost = 24.24 x P x D2 x (1-C) x 24 x 365

  x c/1000 x DUTY x %USE/100 x leak%/100


24.24 is a constant depending on the units of P and D. For pounds per square inch and inches it is 24.24.

P is the absolute pressure.

D is the diameter of the orifice in inches.

C is the proportion of orifice blocked by condensate flow.

24 * 365 converts the hourly steam loss to annual.

c/1000 is the cost of steam per 1000 lb.

leak%/100 is the proportion of the possible leakage.

Annual steam loss is calculated from the cost of steam lost and the probabil-ity of trap failure during a year. The probabil-ity of trap failure is cal-culated from the expected life of the trap. The probability of trap failure each year is prob-ably concentrated around the average (C-100) life C e.g. perhaps a Normal distribution. For a number of such traps, the probabil-ity of failure is cal-culated as

P(trap failure) = 0.5 / estimated life

The annual steam loss is then obtained by multiplying the proba-bility of trap failure by the annual cost of steam lost from the above equation.

Trapo home page, More information, Download files, FAQs, Prices, Links