Introduction

Performing measurements is a process similar to an industrial production process. There are several phases to be observed. In each of these phases specific information is needed. The information is sometimes for the sake of the process itself, however in most cases the information has to be passed to the next phase. Items like purpose, identification, reliability and accuracy have to be addressed and have to be secured. In each phase one must be able to observe the process, see what has been produced, evaluate and visualise results. For simple and incidental processes these actions can be performed manually. As an ad hoc solution data is read into a spreadsheet programm and -when needed- graphs are drawn. When the process becomes more complicated and when a series of measurements has been performed, a more structural solution is needed, where checking and controling the process can be done semi-automatically.

The phases

As a paradigm we distinghuish five phases.

  1. Setup
  2. Measurement
  3. Analysis
  4. Report
  5. Filing

1 Setup is the phase which can be characterised as when your clothes and hands get dirty. A sample or model is prepared, transducers are placed, wired and connected to the data acquisition system. Water is added or drained.

2 Measurement is act of collecting readings of transducers. The information is stored in files.

3 During analysis the results of the previous phase are interpreted. The main activity is deriving new properties, using the recorded data. When trying to understand what the results actualy mean, one has to find out what has happened during the measurement and the preparation of the sample. If a theoretical model is available, results are processed or interpreted according to this model. Even parameters of a theoretical model can be determined. This phase stops when no more information can be derived from the recorded data, using the current knowledge.

4 During the report phase information is retrieved from the archive. It may be reported to a client, it may be placed on a web site, added to an article in a journal, even all information can be transfered to the client. The results of several measurements can be combined and plotted in a graph of written in a table of a PDF document. The main feature is that no new information is generated, the information is merely presented.

5 Filing is the phase were the results of the measurement and the analysis are stored. The results are frozen, and stored in an archive or a database. Maybe some information about storing the file can be added, however this is not the kind of information which adds value for industrial or scientific purposes. Usually the moment in time when the results have been filed and who has filed them is added to the data. 

The process in detail

Even before a data acquisition system is running, we must collect data, in order to secure how we performed an experiment and what the readings mean. The questions we must ask are:

  • Is the experiment done according to the guidelines of a normalisation institute?
  • What (other) procedures have been followed?
  • Which types of transducers are used?
  • Which particular transducers are employed?
  • When were these transducers calibrated?
  • How can we translate the output of a transducer into physical units, like kPa, mm or °C?
  • What is the accuracy of the transducers?
  • What is the physical location (x, y and z position and coordinate system) of the transducers?
  • What kind of sample or material is used?
  • What are its characteristic values?
  • In what way is the sample prepared?
  • What are the ambient conditions (temperature, pressure, humidity, g-level) during the experiment? 

All this information has to be gathered and stored during the measurement phase. Without this information the one that is in charge of the analysis or the reporting of the measurement, has either to guess what has happened or ask the  experimentators "what on earth they have been doing. Sometimes even the method how the results of a measurement are to be analysed is added in the first phase.

Static information, like calibration coefficients, ambient conditions and material properties are added as GEF keywords, basically in the flavor Measurement and Specimen. If any of this information is changing in time, it should be added as variable information in the data section of a GEF file.

Analysis

In the next phase we have to deal with the analysis of the recorded data. For measurements that are commonly performed, usually a norm or a standard exists, issued by organisations like ASTM, BST, DIN, NEN, ISO. If the results of the measurements are to be analysed according to such a standard, we can add this information to the recorded data: (to be) analysed according to NEN ####. This information can be either derived from the way the experiment has been performed or should be added. Not just the type of measurement (permeability), but the particular way (standing or faling head) has to be added as well.

The first stage of analysis comprises the preparation and conditioning of the recorded data. The results of the transducers sometimes have to be translated from volts or counts into physical properties (kPa, mm or °C). The data may be (digitally) filtered, although this is not an obligatory procedure. Finally a selection of the data has to be made. This is either the case if the difference between loading and unloading is to be determined, or when during the measurement the ambient conditions have been altered, or just for determining an initial value.

The second step consists of mathematical operations on the recorded data, like subtracting static components, averaging the signals of several transducers or making least square fits of the recorded signals. E.g. when total stress and pore pressures were measured, the effective stress can be calculated by subtracting both recordings.

In the last step values are determined, which provide rather condensed information about the particular behaviour that the sample has displayed during the measurement, like the thermal expansion coefficient, stiffness, cohesion or permeability of the sample. In experiments where the interaction between structure and sample is investigated, the analysis may also comprise of the displacement or deformation of the structure.

The results of all operations in the analysis have to be added to the data. We emphasize on combining measurement data and analaysis data in one file. The advantage is that whenever doubts rise on the reliability or validity of the analysis, a reanalysis can be done, based on the same data. Furthermore, when the insight in the theoretical background has grown, leading to another way of analysis or when technological advance has led to improved methods, we don't need to perform the measurements again, the data is just waiting to be reused.

Static information, like permeability, cohesion, thermal expansion, can be stored in GEF keywords with the flavour Analysis. Non static information is added as columns in the data block of a GEF file.

Filing

During filing just administrative data is added, stating on which date the results are frozen or stored in a database. Who has stored them? Where in the database are the results stored? The latter is quite important since there should exist only one instance of a measurement. This information is stored as GEF keywords of the Filing flavor.

Report

The type of report one can produce depends on the needs of the client. A consultant or engineer is content with a number, characterising the strength of a material. A researcher might have ideas for a better analysis and so he wants all data. When writing a report for a client, a table of all derived results is needed, and may be graphs of the derived quantities as annexes. The amount of space in a scientific journal imposes yet other needs: a graph with all data, underlining the point or thesis of the article.

The results produced int the report phase differ a lot. A graph may be an image file suitable for a website, a table may be a file suitable for a text document. Sometimes a data or a selection of the data can be transfered to the client. When using standardised software, static information how the data should be (automatically) presented or visualised may be added, als GEF keywords in the Report flavor.

Overview

During the first three phases the amount of information in a GEF file grows. We start with a basic measurement file. The data acquisition system and tailored measuring software provide basic information about the quantities that are recorded. Additional information about transducers, measuring procedures and sample have to be added separatedly. During analysis more information is added. This consists of both static information for derived quantities (like a thermal expansion coefficient) and non-static information, such as columns of translated (physical units) transducers (like effective stress). During analysis the size of a GEF file may double or tripple. Adding filing information does not alter the size of the GEF file significantly, usually its size increases less than 1 kilobyte. When reporting the amount of data usually diminishes. The columns with the raw data in non-physical units are removed, intermediate results are less interesting, a section of the data is selected etc. Some information may be added, however not of scientific or industrial importance, like the number of the annex to be drawn, the line width or line color in a graph.  

Implementation

General 

A successful implementation of this methodology depends on the possibility to perform quality control in each of the four phases. A major advantage of using the same format in all phases, e.g. GEF, is that one can use the same applications and software to visualise results in all phases. Furthermore during analysis one can use the same basic library functions as have been used for the visualisation of GEF data. Reading a file, either measurement, analysis, filing or report, remains the same. The amount of data to be read may vary, the method isn't. Another advantage is that one can develop generic and sometimes highly specialised methods to analyse data. Since these functions can be reused for other projects, the balance between yield and effort can be tipped towards yield due to improved efficiency.

We use GEF for the storage of measurements, their analysis and we use GEF to report results to our clients. A database for filing the GEF files is curently (July 2006) under construction. While the measurements are running, we use software to monitor the signals of the transducers. Furthermore, when clients ask for it, we provide a nearly realtime presentation of the experiment on the World Wide Web, by means of graphs which are derived from the data in the GEF file of the current test. The state of the equipment, like temperatures of bearings, stresses in rotor arm, the air pressure in the setup etc, i.e. quantities which are needed for running the experiment safely, is monitored as well by the same software as for visualising the results of the measurement. This allows for the first and rough quality control of the data. When the data is analysed, new quantities are dervied and added to the GEF file. The same software as used for monitoring the experiment can now be employed, to visualise the derived quantities. Finally, when the client wants to do an analysis of his own, the client can use the same software as we used for monitoring his experiments, and get the first impression of the results. Additionaly tools for the conversion from GEF to other formats, e.g. a Excel spreadsheet, are available.

Examples 

An very nice and convincing example of the way GeoDelft has implemented this methodology for a series of centrifuge tests can be read in this pdf document, This document is made available by the courtesy of Samsung JV. GEF has been used for the purpose of exchanging information between phases and to the client. This implementation has been proven to be successful, since all necessary information of the experiment is either present in the files or can be retrieved, using the information provided in the files.

Another example of a successful implementation can be found in the phases of a series of tests on micro instability, which were preformed in 2004. The document which describes the different phases is not as spectacular as the previously mentioned pdf document, however the accompanying data are not disclosed. Therefore we illustrate the use of phases by means of files of the micro instability test. Results of these tests have been published on the ICPMG conference of 2006. All files of test 5 have been compressed into a ZIP file.

The following actions have been done during the four phases:

  • Measurement. File m_test5.gef contains the registration of the data acquisition system. Additional static data are added. The file contains data of the entire test: from spinning up spinning up, via  the actual test to spnning down.
  • Analysis. In file a_test5.gef the registrations while spinning up and down are removed. The Ng level has been added. Since the Total Stress Transducers TST 1 and TST 2 are not barometrically compensated, their readings are not correct due to the lowering of the air pressure in the centrifuge. They have been corrected for the reduced air pressure. In this phase these correction have been added.
  • Filing. The name of the file, f_test5.gef, the guy that filed this file and the date of filing were added.  
  • Report. In file r_test5.gef some columns have been removed, since they are not interesting for a report, such as the column which provides information about the state of light switches and valves. Furthermore the pressure in the centrifuge has been removed, since it has lost its importance as soons as the total stress tranducers were corrected. Next the column with the rotational speed has been removed; it has been used for the calculation of the Ng level. Finally some static information meant for graphs was added as Reporttexts.
  • No labels

Disclaimer