[Code of Federal Regulations]
[Title 40, Volume 31]
[Revised as of July 1, 2006]
From the U.S. Government Printing Office via GPO Access
[CITE: 40CFR1065.12]
[Page 677-679]
TITLE 40--PROTECTION OF ENVIRONMENT
CHAPTER I--ENVIRONMENTAL PROTECTION AGENCY (CONTINUED)
PART 1065_ENGINE-TESTING PROCEDURES--Table of Contents
Subpart A_Applicability and General Provisions
Sec. 1065.12 Approval of alternate procedures.
(a) To get approval for an alternate procedure under Sec.
1065.10(c), send the Designated Compliance Officer an initial written
request describing the alternate procedure and why you believe it is
equivalent to the specified procedure. We may approve your request based
on this information alone, or, as described in this section, we may ask
you to submit to us in writing supplemental information showing that
your alternate procedure is consistently and reliably at least as
accurate and repeatable as the specified procedure.
(b) We may make our approval under this section conditional upon
meeting other requirements or specifications. We may limit our approval,
for example, to certain time frames, specific duty cycles, or specific
emission standards. Based upon any supplemental information we receive
after our initial approval, we may amend a previously approved alternate
procedure to extend, limit, or discontinue its use. We intend to
publicly announce alternate procedures that we approve.
(c) Although we will make every effort to approve only alternate
procedures that completely meet our requirements, we may revoke our
approval of an alternate procedure if new information shows that it is
significantly not equivalent to the specified procedure.
If we do this, we will grant time to switch to testing using an
allowed procedure, considering the following factors:
(1) The cost, difficulty, and availability to switch to a procedure
that we allow.
(2) The degree to which the alternate procedure affects your ability
to show that your engines comply with all applicable emission standards.
(3) Any relevant factors considered in our initial approval.
[[Page 678]]
(d) If we do not approve your proposed alternate procedure based on
the information in your initial request, we may ask you to send the
following information to fully evaluate your request:
(1) Theoretical basis. Give a brief technical description explaining
why you believe the proposed alternate procedure should result in
emission measurements equivalent to those using the specified procedure.
You may include equations, figures, and references. You should consider
the full range of parameters that may affect equivalence. For example,
for a request to use a different NOX measurement procedure,
you should theoretically relate the alternate detection principle to the
specified detection principle over the expected concentration ranges for
NO, NO2, and interference gases. For a request to use a
different PM measurement procedure, you should explain the principles by
which the alternate procedure quantifies particulate mass similarly to
the specified procedures. For any proportioning or integrating
procedure, such as a partial-flow dilution system, you should compare
the alternate procedure's theoretical response to the expected response
of the specified procedures.
(2) Technical description. Describe briefly any hardware or software
needed to perform the alternate procedure. You may include dimensioned
drawings, flowcharts, schematics, and component specifications. Explain
any necessary calculations or other data manipulation.
(3) Procedure execution. Describe briefly how to perform the
alternate procedure and recommend a level of training an operator should
have to achieve acceptable results.
Summarize the installation, calibration, operation, and maintenance
procedures in a step-by-step format. Describe how any calibration is
performed using NIST-traceable standards or other similar standards we
approve. Calibration must be specified by using known quantities and
must not be specified as a comparison with other allowed procedures.
(4) Data-collection techniques. Compare measured emission results
using the proposed alternate procedure and the specified procedure, as
follows:
(i) Both procedures must be calibrated independently to NIST-
traceable standards or to other similar standards we approve.
(ii) Include measured emission results from all applicable duty
cycles. Measured emission results should show that the test engine meets
all applicable emission standards according to specified procedures.
(iii) Use statistical methods to evaluate the emission measurements,
such as those described in paragraph (e) of this section.
(e) We may give you specific directions regarding methods for
statistical analysis, or we may approve other methods that you propose.
Absent any other directions from us, use a t-test and an F-test
calculated according to Sec. 1065.602 to evaluate whether your proposed
alternate procedure is equivalent to the specified procedure. We
recommend that you consult a statistician if you are unfamiliar with
these statistical tests. Perform the tests as follows:
(1) Repeat measurements for all applicable duty cycles at least
seven times for each procedure. You may use laboratory duty cycles to
evaluate field-testing procedures.
Be sure to include all available results to evaluate the precision
and accuracy of the proposed alternate procedure, as described in Sec.
1065.2.
(2) Demonstrate the accuracy of the proposed alternate procedure by
showing that it passes a two-sided t-test. Use an unpaired t-test,
unless you show that a paired t-test is appropriate under both of the
following provisions:
(i) For paired data, the population of the paired differences from
which you sampled paired differences must be independent. That is, the
probability of any given value of one paired difference is unchanged by
knowledge of the value of another paired difference. For example, your
paired data would violate this requirement if your series of paired
differences showed a distinct increase or decrease that was dependent on
the time at which they were sampled.
[[Page 679]]
(ii) For paired data, the population of paired differences from
which you sampled the paired differences must have a normal (i.e.,
Gaussian) distribution. If the population of paired difference is not
normally distributed, consult a statistician for a more appropriate
statistical test, which may include transforming the data with a
mathematical function or using some kind of non-parametric test.
(3) Show that t is less than the critical t value, tcrit, tabulated
in Sec. 1065.602, for the following confidence intervals:
(i) 90% for a proposed alternate procedure for laboratory testing.
(ii) 95% for a proposed alternate procedure for field testing.
(4) Demonstrate the precision of the proposed alternate procedure by
showing that it passes an F-test. Use a set of at least seven samples
from the reference procedure and a set of at least seven samples from
the alternate procedure to perform an F-test. The sets must meet the
following requirements:
(i) Within each set, the values must be independent. That is, the
probability of any given value in a set must be unchanged by knowledge
of another value in that set. For example, your data would violate this
requirement if a set showed a distinct increase or decrease that was
dependent upon the time at which they were sampled.
(ii) For each set, the population of values from which you sampled
must have a normal (i.e., Gaussian) distribution. If the population of
values is not normally distributed, consult a statistician for a more
appropriate statistical test, which may include transforming the data
with a mathematical function or using some kind of non-parametric test.
(iii) The two sets must be independent of each other. That is, the
probability of any given value in one set must be unchanged by knowledge
of another value in the other set. For example, your data would violate
this requirement if one value in a set showed a distinct increase or
decrease that was dependent upon a value in the other set. Note that a
trend of emission changes from an engine would not violate this
requirement.
(iv) If you collect paired data for the paired t-test in paragraph
(e)(2) in this section, use caution when selecting sets from paired data
for the F-test. If you do this, select sets that do not mask the
precision of the measurement procedure. We recommend selecting such sets
only from data collected using the same engine, measurement instruments,
and test cycle.
(5) Show that F is less than the critical F value, Fcrit, tabulated
in Sec. 1065.602. If you have several F-test results from several sets
of data, show that the mean F-test value is less than the mean critical
F value for all the sets. Evaluate Fcrit, based on the following
confidence intervals:
(i) 90% for a proposed alternate procedure for laboratory testing.
(ii) 95% for a proposed alternate procedure for field testing.