GISdevelopment.net ---> GITA 1997 ---> Fundamental & Economic Issues of AM/FM/GIS

Life Cycle Approach to Managing an AM/FM/GIS Project

Jay Stinson, PMP
Intergraph Corporation
Huntsville, Alabama 35894-0001 (LR24B1)


Introduction
It can be argued that the most important AM/FM investment is in securing cogent AM/FM data. Hence, it is vital that the AM/FM/GIS client be aware of powerful techniques for assuring that they are receiving the highest quality data that their project budget can secure.

Herein, the reader will find a brief discussion of several issues and concepts that could be applied toward measuring the cogency of AM/FM/GIS data. Armed with these concepts and techniques, the data provider and client can both derive objective statistics about whether or not the data meets predefine criteria. These techniques are more than a means for policing the work of the data provider, for the data provider has an equal opportunity to apply these concepts in-house, before the product is examined by an agent of the client.

This paper does not include a systematic work flow procedure than can be extracted and subsequently adopted for use in a project. Aside from the fact that the size of this work cannot accommodate such an exhaustive presentation, each AM/FM/GIS endeavor has its own unique characteristics that together allude to advantages and disadvantages of adopting the solutions of other endeavors. Instead, the concepts 360?presented herein are available for strong consideration while preparing a QA/QC solution for evaluating the cogency of AM/FM/GIS data product.

Generally, the task is securing the means for the AM/FM/GIS client to objectively determine whether or not data product is of high enough quality to be acceptable for use in their AM/FM/GIS solution. The overall strategy is to find the least expensive means for rejecting a deliverable unit (DU) within the context of project policies and agreements between the conversion vendor and the Client.

In the interest of maintaining a cost-effective QA/QC effort, the suggested strategy is to reserve the more expensive, labor-intensive inspection procedures to the very end, whereby a DU would have already been vigorously tested using automated tools or by exposing the DU’S to inspection types whose rejection vmuld render the DU ineligible for further inspection.

It can be said that there are two levels of QA/QC: (1) Digital Integrity, and (2) Information Accuracy. Digital Integrity refers to whether or not the data has been placed and presented according to established rules for providing digital recognition and digital relationships. Information Accuracy refers to whether or not that data represents true information as found in the source records. The scope of this paper is limited to techniques associated with Information Accuracy QA/QC procedures that are to be applied against data product that has already cleared through all Digital Integrity inspection procedures.

Information Accuracy QA/QC procedures could be established according to the principles described in American Nationa/ Standard Sampling Procedures and Tab/es for hspection by Attributes, Zf.4-f993, as prepared and maintained by the American Society for Quality Control Standards Committee. This document “implies a consensus of those substantially concerned with its scope and provisions, ... intended as a guide to aid the manufacturer, the consumer, and the general public” (ANS1/ASQC ZI .4-1993, title page) and is intended for use by any organization that is able to apply its concepts.

The ANSt/ASQC standards provide a solid foundation for building a QA/QC operation that grants the client the opportunity to actively monitor and control the quality of their incoming conversion data, and lends objective authority to procedures derived therefrom, affording the client even greater power in policing the quality of their AM/FM/GIS solution’s data product.

Overview of ansuasqc-based random sampling concepts
Discussing AM/FM/GIS data in the context of ANS1/ASQC standards requires that we define units of inspection. For the purposes of this paper, we shall say that the data provider ships a set of deliverable units (DU’S) to the AM/FM/GIS client (the Client). Generally a DU, usefully defined, would be a collection of map areas that would contain an average of 50 to 3000 “features”. Each feature could be a single object such as a segment of uniquely identifiable gas pipe, a secondary power line, or a specific valve, etc. 361?Often, these features will have text attributes associated with them - it is acceptable to consider these text attributes as part of their respective feature.

Hence, using these definitions, the data provider ships a set of DU’S to the AM/FM/GIS client, with each of these DU’S containing several hundred features. These features are subsequently examined and evaluated in determining whether their respective DU should be accepted and incorporated as part of the AM/FM/GIS data solution.

In order to be able to retain the authority of the ANS1/ASQC standards, it is imperative that certain constraints be respected while developing a set of QA/QC work flow procedures. There are four such constraints that are instrumental in developing QA/QC work flow procedures such that those procedures retain the authority of the ANS1/ASQC standards:
  1. the order in which the DU’S are to be inspected -it is important that the order in which DU’S are inspected be random.
  2. how many features to inspect - determined according to the quantity of features in each DU as well as the prescribed inspection level intensity,
  3. techniques for randomly identifying those features that are to be inspected - a fair inspection requires randomly choosing a sample of features that are to be examined.
  4. objective descriptions of the nature of each detected error - this objectivity provides a means for reporting errors in a controlled QA/QC environment.
These four considerations are described as follows:
  1. ldentifying the Sequence Order for Inspecting DU’S During Information Accuracy Inspection Procedures, it is important that the order in which DU’S are inspected be random. This requirement stems from the fact that it is necessary to track inspection trends such as too many failures as well as positive trends of conversion excellence. These trends provide the means for authorizing more relaxed or more stringent QA/QC operations.

    As an example of how tracking inspection trends is important to the overall inspection flow, the Responsible Authority (RA) can decide whether or not to shift to a more relaxed level of inspection activity in cases where ten consecutively inspected DU’S have passed inspection. The random ordering of the DU’S assures that such trends would be fairly derived, with the conversion vendor being unable to predict the order in which the DU’S would be subject to inspection.

    To prepare the Inspection Queue, prepare a list of DU’S that are to be inspected such that the resulting list is a random ordering of the original list of 362?DU’S. This “randomized Inspection Queue shall hereafter serve as the official Inspection Queue for this inspection cycle.

  2. Determining How Many Features to Inspect The most thorough method for checking information accuracy is to check each and every feature that is delivered - however, this method is cost prohibitive in that each and every feature must be examined and justified by the inspector (a technique similar to establishing a second, parallel conversion effort). Hence, the challenge of implementing information accuracy procedures is to determine how many features must be examined, and how many errors can be tolerated, before an inspector can make a confident recommendation to accept or reject the DU.

    The ANSi/ASQC standards provide a means for identifying the quantity of features that must be checked. This figure is derived from (1) the quantity of total features for a DU, and (2) the current level of inspection intensity, as well as the mandated level of data accuracy (set at the project level, usually as a percentage, such as 95% or 99%).

    1. Quantity of Delivered Features The quantity of features is a significant indicator as to how many features will require examination before the inspector can make an informed judgement about the overall quality of the DU. This quantity of inspection features is not a direct percentage of the overall count of features in the DU. Instead, the smaller the total count of features, the greater the actual percentage of features that will be examined. The exact amount of features that must be inspected in provided in a table in Arneticar? iVationa/ Standard Sampling Procedures and Tables fbr Inspection by Attributes, Z7.4-7993. This table also defines the amount of permissible errors, thereby identifying how many detected errors would be required to render the DU as unacceptable.

      The number of sample [features] inspected shall be equal to the sample size given by the plan. If the number of nonconforming [features] found in the sample is equal to or less than the acceptance number, the [DU] shall be considered acceptable. If the number of nonconforming [features] is equal to or greater than the rejection number, the [DU] shall be considered not acceptable. (ANSI ZI.4 p.6, paragraph. 10.1.1, italics added) The acceptance number k a prescribed value according to the ANSI ZI.4 sampling plans. As an example, should a DU contain 2,750 features, under a normal sampling plan, 125 features would be selected from throughout the entire DU and inspected. Given a 99?40acceptance criteria, identifying three (3) erroneous features wmuld render a recommendation to (nevertheless) accept the entire DU, while identifying four (4) erroneous features would render a recommendation to reject the entire DU.

      The client’s ANSl/ASQC-based sampling plan should accept the great majority of the DU’S that the conversion vendor submits, provided the nonconformities per hundred units in these DU’S be no greater than one (1) unit. This requirement establishes the AQL level as 1.0, and can also be referred to as an “Acceptance Criteria of 99%”.

    2. Level of Intensity The level of intensity refers to whether the DU is to be inspected via a more relaxed or more stringent inspection criteria, identified according to whether more or less features are to be examined. This so-called level of intensity is a system for allowing an inspector to reduce inspection costs for product that is consistently of highest quality, as well as allowing an inspector to devote more resources to examine product that is consistently below criteria.

      In general, an example DU of 2,750 units having a 99°A acceptance criteria, would render the following sampling plans, each according to their respective inspection level:

      • Normal: inspect 125 features, three (3) errors allowed.
      • Relaxed: inspect 20 features, zero (0) errors allowed, one(1) error allowed although normal inspection is reinstituted.
      • Tiqhtened: inspect 200 features, three (3) errors allowed.

      Prior to releasing a set of DU’S for inspection, the RA identifies the level of intensity for each of the DU’S that are to be inspected, based on inspection trends. ANSI ZI.4 sampling plans include controlled provisions for changing the intensity of inspection activity throughout the life of the QA/QC operation. These controlled provisions, based on the performance trends of the current inspection queue, indicate when a plan should be switched from normal to tightened and normal to relaxed, as well as tightened to normal and relaxed to normal. The switching rules are summarized, based on ANSI Z1.4 (pages 4 and 5), as follows (refer to said document for authoritative details):

      • From Normal to Tightened (poor trend): any two (2) out of five (5) consecutive DU’S have been found to be non-acceptable, after being inspected at a normal level of intensity.

      • From Tightened to Normal (recovering trend): five (5) consecutive DU’S have been found to be acceptable after being inspected at a “tightened” level of inspection intensity.

      • From Normal to Relaxed (excellent trend): four (4) conditions are met, including (a) preceding ten (1 O) DU’S have been found to be acceptable, (b) preceding ten (1 O) DU’S were identified as having a significantly less quantity of errors than is allowed (per applicable tables), (c) production is at a steady rate, and (d) relaxed inspection is considered desirable by the RA.

      • From Relaxed to Normal (normalizing trend): any one of four (4) conditions are encountered while in relaxed level of inspection, including (a) a DU is rejected, or (b) during inspection, the sampling procedure may terminate without reaching a decision (DU passes, although it “almost failed” per applicable tables), or (c) production becomes irregular or delayed, or (d) other conditions warrant a switch to normal inspection, as determined by the RA. /li>

    Switching the inspection plan lends significant power to the RA in enhancing their confidence in the acceptability of the overall project data product, while taking advantage of opportunities to minimize inspection labor at times when the data is judged to be of excellent quality. Once the Inspection Level is assigned for a DU, that inspection level should be maintained throughout the DU’S current inspection cycle.

  3. Identifying Sequence of Specific Features to be Inspected Once the inspector knows the quantity of features that is to be inspected, they must randomly select the features that are to be inspected. The random targeting of features assures all parties that there is no bias toward obviously correct features nor toward obviously defective features (although the ANSi/ASQC standards allow for the client to mark obviously erroneous features for subsequent repair by the data provider).

    In order to be able to demonstrate complete objectivity in choosing the particular features, it is suggested that objective procedures be used in randomly targeting the features. Two objective procedures are as follows:

    1. Place a Cartesian coordinate grid over the DU map area, and by using randomly generated numbers, derive the X,Y coordinates of the features that are eligible for examination.

    2. Develop and use a software tool that is capable of randomly selecting a specified quantity of features directly from the digital data file.

  4. Objectively Defining Errors While this requirement is not derived from directly ANS1/ASQC standards, it is important to be able to specify, objectively, what is wrong with any defective features. In most industries, it is readily obvious how to describe an error with a tangible product. However, with the product being AM/FM/GIS data features, the nature of the information error is not as readily discernible. Hence, a suggestion for rectifying this challenge is for each feature to be described according to the three following error characteristics:

    • Completeness: whether or not the feature is extra, or whether a feature is missing from the conversion data.
    • Layout/Position: refers explicitly to the position of features with respect to other features.
    • Accuracy: refers to whether or not a feature portrays correct information, usually refers to an attribute of a feature such as a diameter, vintage, status, etc.
Any feature that is inspected and found to be in error for any of these three reasons should be counted as a single erroneous feature.

Conclusion
The techniques described herein illustrate many ways in which an AM/FM/GIS client can establish an objective environment for discussing data quality issues with the data provider. Adopting the techniques described in Arneficar? /Vafiona/ Standard Sampling Procedures and Tables for Inspection by Attributes, Z1. 4-1993 lends objective authority to the client in evaluating whether or not to accept contractually delivered data product. Indeed, these standards include a mechanism whereby the client can refuse any further deliveries until such time the data provider demonstrates that it has taken corrective action in its data production operation. Ironically, the greatest benefit of this type of a QA/QC environment is that these techniques can also be applied by the assure that the quality of their product own business standards.

© GISdevelopment.net. All rights reserved.