You are on page 1of 3

Guest Column

Robert X. Perez, Celanese Chemicals

MTBF Calculations
in a Nutshell

M
ean time between failures (MTBF) is a What is a failure?
useful metric when failures are randomly First, define a failure. This definition varies
distributed — a good assumption for widely among users. I define a failure as any event
most complex machines or devices.
requiring the replacement of a machine component
For example, because pumps have numerous
due to a loss of equipment function. This definition
failure modes, i.e. seal, bearing, and coupling fail-
excludes all work related to preventative mainte-
ures, their failure distributions tend to be random
or exponential. For this reason, the MTBF metric is nance, adjustments, or improvements.
useful in industries operating complex devices such
as pumps, motors, engines, turbines, trucks, etc. Single Component Testing
Most users of machinery have complex popula- Let’s assume we run a single new component in
tions (or fleets) of equipment. A fleet can be com- a controlled test for a time T. Assume that when it
prised of machinery of different sizes, manufactur- fails, it is replaced with a new or rebuilt component
ers, and services. Equipment can also be designated instantaneously. This means we assume the MTTR
as spared and unspared. to be zero. If the component fails F times, the aver-
Users typically calculate and trend the aggre- age mean time between failures (mtbf) is deter-
gate MTBF of their fleet to handle this complexity. mined with the equation:
Although frequent and lively discussions may
debate exactly how this should be done, most com- mtbf = T ÷ F (1)
panies actually use similar equations for this calcu-
Since we assumed the MTTR to be zero, we can
lation. I want to explain the most common equa-
tions for determining the fleet MTBF, with some say the reciprocal of mtbf is equal to the average failure
examples. rate (λ) for the testing interval T, or 1 ÷ mtbf = λ.
One problem of great interest is determining
which pieces of equipment in the fleet are “bad Fleet Calculations
actors.” Let’s walk through how the in-service For a mix of equipment types, such as pumps,
MTBF is determined using real-world data, and trucks, etc., users typically calculate the aggregate
how this data is used to select “bad actors” from a mean time between failures of the fleet (MTBFfleet)
complex population of equipment. with this equation:

Definitions MTBRfleet = N T ÷ F (2)


T = test time (in days or years)
Example 1:
N = number of components in the population or fleet
M = number of components operating at any given time 500 pumps operate for a year, experiencing 100
F = total failures in the population during testing time (T) failures. What is the average MTBF of the fleet?
MTBFfleet = 500 x 1 year ÷ 100 failures = 5 years
%OS = Percent of the time on-stream between failures
OST = On-stream time = Percent time on-stream x time
= %OS x T If it costs $4000 to repair a pump, what is the fore-
λ = failure rate (failures/unit time) casted annual repair cost for this pump population?
MTTR = mean time to repair $4000 x 100 per year = $400,000 per year

6 AUGUST 2005 www.pump-zone.com PUMPS & SYSTEMS


Aug05PUMPS&SYSp6-7 7/20/05 10:09 AM Page 7

In-service Calculations M = 50 + 225 pumps


If some machines in the population are idle M ÷ N = 275 ÷ 500 = 55%
during the testing time, then a formula different MTBFin-service = (275 pumps x 1 year) ÷ 100 failures
from equation (2) is required to determine the aver- = 2.75 years/failure
age MTBF for machines that are in service
This represents a λ of 0.36 failures/year.
(MTBFin-service). If only the fraction M ÷ N of the
machines are operating during the testing time, CONTINUED on PAGE 94…
then equation (1) becomes:
MTBFin-service = N (M ÷ N) T ÷ F (3)
Simplified, this becomes:
MTBFin-service = M T ÷ F (4)

➥Note 1
To be more accurate, T should be the
actual mission time of the in-service
population. So, if the process is on-
stream 90 percent of the time, then
Tactual = OST = 0.9 x T. This is
written as:
MTBFin-service = M (%OS x T) ÷ F =
M x OST ÷ F (5)

Example 2:
500 pumps operate for a year,
experiencing 100 failures. If only
50 percent of these pumps are
operating at one time, what is the
average in-service MTBF?
MTBFin-service = (500 x 0.5 x 1 year)
÷ 100 failures = 2.5 year/failure
This represents an in-service fail-
ure rate (λ) of 0.4 failures per year.

➥Note 2
Here we will assume λ (failure rate)
= 1 ÷ MTBF by saying that MTBF
> MTTR. This is usually a good
approximation. For better accuracy,
state that λ = 1 ÷ (MTBF –
MTTR) if you know the MTTR
Circle 122 on Reader Service Card

value.

Example 3:
500 pumps operate for a year, ex-
periencing 100 failures. Review of
the operations data finds that 10
percent of the pumps are unspared
and the rest are spared. What is
the average in-service MTBF?

PUMPS & SYSTEMS www.pump-zone.com AUGUST 2005 7


Aug05PUMPS&SYSp88-94 7/20/05 3:02 PM Page 96

CONTINUED from PAGE 7…

Example 4: Example 5:
There are three pumps in a single service. Only two Consider an example where you only have 2 years of
are required for the process. If these pumps experi- failure data for two spared pumps. During this time
enced 5 failures last year when they were operated you recorded one failure (Assume M = 1, %OS =
only 90 percent of the time, what is the average in- 98%).
service MTBF? MTBFin-service (apparent) = (1 x 0.98 x 2 years) ÷ 1
MTBFin-service = 2 pumps x 0.9 year ÷ 5 failures = = 1.96 years/failure
0.36 years/failure MTBFin-service (lower limit) = (1 x 0.98 x 2 years) ÷ 2
This represents a λ of 2.78 failures/year. = 0.98 years/failure

Calculating MTBF with limited Conclusions


field data Fleet data forecast annual repair costs and when
In cases where field data is limited, equation (5) the next fleet failure will occur, as shown in Example
should be modified to bracket the possible MTBF 1. In-service data determine actual in-service failure
range as: rates, which help in reliability modeling, assessing
component reliability, and identifying “bad actors.”
MTBFin-service (apparent) = M (%OS x T ) ÷ F = For example, if the fleet MTBF is 4 years/failure,
M x OST ÷ F but a particular set of spared pumps has an average
MTBFin-service (lower limit) = M (%OS x T ) in-service MTBF of 0.5 years/failure, this probably
÷ (F + 1) = M x OST ÷ (F + 1) indicates a pump design or operational problem.
P&S
This accounts for the possibility that a failure
occurred one day, hour, or second before you start- Contributing editor Robert X. Perez is a staff engineer
ed recording data and one failure occurred soon of rotating equipment at Celanese Chemicals. He can
after the evaluation time ended. Once F ÷ (F + 1) be reached at 361-584-6936 or rperez@bishopfacili-
approaches 1, the distinction between the apparent ty.com.
and lower limit values becomes less significant.
Pumps & Systems Advertiser’s Index
Advanced Sealing International . . . . . . .49 Graphite Metallizing Corporation . . . . .77 ROC Carbon . . . . . . . . . . . . . . . . . . . . . .96
Anvil International, Inc. . . . . . . . . . . . . .46 Griffin Dewatering Corporation . . . . . . .56 Schlumberger . . . . . . . . . . . . . . . . . . . . . .3
Arkema, Inc. . . . . . . . . . . . . . . . . . .22, 23 Hydro, Inc. . . . . . . . . . . . . . . . . . . . . . . .28 Sealing Equipment Products Co. Inc. . . .65
Baker Manufacturing Company / Haight Hydromatic Pump . . . . . . . . . . . . . . . . .48 seepex, Inc. . . . . . . . . . . . . . . . . . . . . . .66
Pump Division . . . . . . . . . . . . . . . . . . . .IBC InduMar Products, Inc. . . . . . . . . . . . . . .85 Serfilco, Ltd. . . . . . . . . . . . . . . . . . . . . . .86
Baldor Electric Company . . . . . . . . . . . .25 Inpro/Seal Company . . . . . . . . . . . .41, BC Sims Pump Valve Co. . . . . . . . . . . . . . . .96
Bently Pressurized Bearings . . . . . . . . . . . .7 ISG Thermal Systems USA, Inc. . . . . . . .57 Sims Pump Valve Company Inc. . . . . . .67
Boerger, LLC . . . . . . . . . . . . . . . . . . . . . .50 John Crane, Inc. . . . . . . . . . . . . . . . .30, 31 SKF Sealing Solutions . . . . . . . . . . . .42, 43
CEC Vibration . . . . . . . . . . . . . . . . . . . . .95 KTR Corporation . . . . . . . . . . . . . . . . . .58 Sponsler, Inc. . . . . . . . . . . . . . . . . . . . . .87
Cole-Parmer Instrument Company . . . .51 Load Controls . . . . . . . . . . . . . . . . . . . . .95 Standard Alloys . . . . . . . . . . . . . . . . . . .68
Continental Electric Motors, Inc. . . . . . .75 Lutz-JESCO America Corporation . . . . .78 Sulzer . . . . . . . . . . . . . . . . . . . . . . . . . . .33
Dan Bolen . . . . . . . . . . . . . . . . . . . . . . . .96 Metallized Carbon Corporation . . . . . . .59 SyncroFlo, Inc. . . . . . . . . . . . . . . . . . . . .69
Des Case . . . . . . . . . . . . . . . . . . . . . . . . .95 Met-Pro Corporation/Dean Pump Div. . .34 Tarby . . . . . . . . . . . . . . . . . . . . . . . . . . . .82
Diamond Innovations . . . . . . . . . . . . . . .76 Met-Pro Corporation/Fybroc Div. . . . . . .34 Thordon Bearings Inc. . . . . . . . . . . . . . .70
Dickow Pump Company . . . . . . . . . . . .52 Moyno, Inc. . . . . . . . . . . . . . . . . . . . . . .IFC Trico Manufacturing Corporation . . . . .71
Dupont Vespel . . . . . . . . . . . . . . . . . . . .83 MSE of Canada . . . . . . . . . . . . . . . . . . . .95 Tuf Lok . . . . . . . . . . . . . . . . . . . . . . . . . . .95
DXP Enterprises, Inc. . . . . . . . . . . . . . . .53 National Oilwell Varco . . . . . . . . . . . . . .45 Vaughan Company, Inc. . . . . . . . . .36, 37
EagleBurgmann Industries LP . . . . . . . . .54 National Tribology Services, Inc. . . . . . . .86 Vertiflo . . . . . . . . . . . . . . . . . . . . . . . . . . .96
Ebac . . . . . . . . . . . . . . . . . . . . . . . . . . . . .96 Netzsch Inc. . . . . . . . . . . . . . . . . . . . . . .61 Viking Pump, Inc. . . . . . . . . . . . . . . . . . .39
Eccentric Pumps, LLC . . . . . . . . . . . . . . .83 NIKKISO Pumps America, Inc. . . . . . . . .60 Water Works . . . . . . . . . . . . . . . . . . . . . .95
Emotron Inc. . . . . . . . . . . . . . . . . . . . . . .84 Oberdorfer Pumps . . . . . . . . . . . . . . . .62 Wilden Pump & Engineering Co. . . . . . .72
F. E. Myers . . . . . . . . . . . . . . . . . . . . . . .47 Orival, Inc. . . . . . . . . . . . . . . . . . . . . . . .79 Worldwide Electric Corporation . . . . . . .73
Falk . . . . . . . . . . . . . . . . . . . . . . . . . . . . .95 Parametrix . . . . . . . . . . . . . . . . . . . . . . . .96 Yamada Corporation . . . . . . . . . . . . . . .74
Flux Pumps Corporation . . . . . . . . . . . .84 Peerless Pump Company, Inc. . . . . . . . .63
Frontline Industries, Inc. . . . . . . . . . . . . .85 Periflo Inc. . . . . . . . . . . . . . . . . . . . . . . .80 *Ad index is furnished as a courtesty and
Garlock Sealing Technologies . . . . . .26,27 Polaris Pumps, LLC . . . . . . . . . . . . . . . . .64 no responsibility is assumed for incor-
Godwin Pumps . . . . . . . . . . . . . . . . . . .55 Pumping Solutions . . . . . . . . . . . . . . . . .81 rect information.

94 AUGUST 2005 www.pump-zone.com PUMPS & SYSTEMS

You might also like