3. CUSUM ( cumulative sum control chart) (Quality
Control):
There is no previous definition.
Definition:
Objective monitoring
or quality control has been used for surveillance: methods have been developed to
detect increases in rare events, such as birth malformations and to detect
minimal epidemics. A review of the concepts and definitions of quality control procedures
that are commonly used in clinical chemistry have also been presented.The use of
the cumulative sum (cusum) has been suggested for both surveillance and quality
control. Its use for examining sequential measures or for looking for changes over
time has recently been described. It has also been used for plotting temperature
charts for assessing antimicrobial treatment in neutropenic patients.
Because the cusum
shows changes over time it can be used by individual practitioners to monitor
their own performance as a form of quality control to give proof of ongoing competence
in a particular skill. It can also be used to show progress in mastering a new technique.
One advantage of this sort of self assessment is that an acceptable level of
attainment must be defined so that how well it is met can be quantified. It has
been suggested, for instance, that the completion rate for colonoscopic examination
should be above 90%.
The cusum has
been described as a useful graphical tool for discerning trends. For a series
of observations XI, X2, . . . Xn the cusum can
be defined as
Sn = Σ(Xi-X0)
where Xi=
0 for a success and Xi= 1 for a failure. Xo is a reference
or target value. The cusum is interpreted by looking at its gradient or slope. For
a series of observations which meet the specified criterion the cusum should be
relatively level. If the reference value is specified in terms of an acceptable
failure rate, for instance, the cusum will increase and consequently will have a
positive slope when the failure rate is higher than the target value, and it
will decrease when the failure rate is lower than that specified. This means that
if the failure rate were specified as 10%, for instance, 0 9 would be added to the
cusum for a failure and -0.1 for a success. In a series of examinations consisting
of a success followed by a failure and four successes the cusum would take the
values -0.1, 0.8, 0.7, 0.6, and 0.5.
EXAMPLE
The first example
concerns the outcome of the 361 examinations done by the experienced
colonoscopist; the cusum is shown in figure 1. As it had been suggested that an
acceptable failure rate for colonoscopies should be less than 10% this value
was used as the reference or target value for the cusum. As a failure rate of
20% was considered unacceptable boundary lines were constructed to identify
series of observations in which the failure rate increased to 20% or more. It was
found that, with a small amount of rounding, a boundary line 2.5 units above
the horizontal axis provided a small risk of accepting an unsatisfactory series
of observations and of rejecting a series of observations that were satisfactory.
By drawing a series of lines 2.5 units apart it was possible to monitor the overall
performance of the colonoscopist; the inter-section of any of these lines from below
marked a period that included more failures than expected. Where this happened the
boundary line became the new target line.
The cusum cut
the boundary line for the first time after 27 colonoscopies, which included six
failures. The overall failure rate for these was 22%. The cusum intersected
with the second boundary line after a further 38 examinations, which included
six failures (failure rate 16%). It cut the same boundary line again at the
164th observation, the preceding four observations having included three
failures. Some sections of the graph represent long runs of successes, the longest
of which was 43. After such a long run the target line should be recentered by
moving it to the one below.
It is
impossible to tell from a retrospective analysis whether the alarms were the
consequence of poor operator performance, an unusual run of technically difficult
colonoscopies, or instrument malfunction. The control chart is especially
sensitive to short runs of failures, and identified them quickly. It could be argued
that this was a shortcoming as in a retrospective analysis it can be seen these
short runs were of little consequence. Had the failures persisted, however, it would
be an advantage to identify such periods as soon as possible.
(Williams, S. M., et al, Quality Control: an application of cusum, p. 1359)
(Kosanke,
K., et al, CIMOSA: enterprise
engineering and integration)
4. CIMOSA (Computer Integrated Manufacturing Open System Architecture) (Modelling)
There
is no previous definition.
Definition:
CIMOSA (CIM Open System Architecture) provides a
process oriented modelling concept that captures both the process functionality
and the process behaviour (Fig. 3). It supports evolutionary
enterprise modelling, e.g., the modelling of individual enterprise domains (DM)
which may contain one or several individual processes (P-1, P-2, …). Domains
and processes are defined by the user according to his/her needs for controlling
the business operations. Processes themselves should be defined as
significantly large pieces of functionality which produce a certain end-result
for a defined customer. Customers may be internal or external to the
enterprise.
Fig. 3
CIMOSA always models the relations to
the internal and external environment (in terms of events and results). This
allows models to be integrated with other process models at a later point in
time. The relations will become the links to the added models.
To handle complexity, CIMOSA follows an
enterprise engineering concept which separates functionality (EA=Enterprise
Activity) and behaviour (BRS=Behavioural Rule Set) allowing to change one
without having to change the other.
Large processes are broken down into
smaller ones ending in networks of enterprise activities which are connected by
the behavioural rule sets. It is this network of enterprise activities which
represent the business process model to be used in the operational support.
Processes are triggered by events (e.g.,
customer order arrival, timer, etc.) and completed by producing their end
result(s). Producing the end-result may start another process (e.g., shipment)
or be used to synchronise other processes. Processes can start one another
demanding sub-results to be produced, which are used in the course of their own
processing. The calling process may have to go into a wait state or may
pre-schedule the sub-result.
Fig. 4 shows a simplified
example of the CIMOSA modelling methodology. Having determined the business
domain to be modelled (e.g., order processing) and its relationships with its
environment (e.g., customer, supplier), the individual, but communicating business
processes (P-1/3) and their activities (for P-2=EA1/4) are identified.
Fig. 4
The information items used in the model
are identified as inputs and outputs of the enterprise activities (Fig. 4, lower part for EA4 only). The
inputs can define the things to be processed (material/parts, information), the
resources needed for the processing and control information for the processing
by the particular activity. Outputs will be the result and the ending statuses
of both the activity and the resources. The information attached to the ending
status may be used in monitoring processes for administrative purposes or for
exception handling. Inputs and outputs are aspects of enterprise objects which
are represented in the information part of the enterprise model.
The BRS identifies the conditions under
which the different activities will be started. Business processes are started
by events (e.g., orders), however, the actual start activity may be different
for different events (e.g., Shop Floor Order a⇒EA1 and Shop Floor
Order b⇒EA3). Process results may also be
produced by different activities and at different times (EA2⇒Product a and EA4⇒Product b). Events and results will be
exchanged with external partners (customers or suppliers) or between different
business processes.
No comments:
Post a Comment