LITERATURE CITED i.
2. 3.
M. A. Zemel'man et al., Coll. of Reports of the 6th All-Union Scientific--Technical Conf. on "Promising developmental trends in electronic instrument design," All-Union ScientificResearch Institute of Electrical Instrumentation, Leningrad (1985). E. I. Girls and E. A. Piskulov, Analog--Digital Converters [in Russian], Energoizdat, Moscow (1981). V. N. Ivanov and E. I. Tsvetkov, Prib. Sistemy Upr., No. 5, 20 (1984).
In the Spring of 1986, The Third All-Union Conference on Theoretical Metrology was held in Leningrad. The papers dealt with researches on measurement theory, particularly the tightening requirements for measurement practices, including improved methodological principles and concept systems, and developments in the theory of systems for ensuring unified measurements. The conference confirmed the importance of topics related to more complicated and automated measurements, the use of computing, and the general introduction of data-acquisltion suites and logging systems. We present here the first part of a selection of papers that attracted the main interest at the conference.
GNOSEOLOGICAL PRINCIPLES AND BASIC METROLOGICAL CONCEPTS V. A. Granovskii, L. M. Gutner, L. I. Dovbeta, and V. V. Lyachnev
UDC 001.8:389.14
Metrology as a science has arisen from accumulated knowledge on measurements and methods of unifying them. This body of knowledge became a science via the creation of a system of basic concepts and fundamental principles. Attempts have been made to formulate such a system in measurement theory [i, 2]. This now needs extension and a firmer basis. Several papers in this journal have dealt with basic measurement-theory concepts. However, these have dealt only with individual measurement elements or with research as a whole where measurement is a basic procedure. Correspondingly, the conclusions cannot give an integral concept on measurements. It is generally accepted that measurement is a cognitive procedure, so initial measurement-theory concepts are essentially related to knowledge-theory ones. Initial concepts in knowledge theory (epistemology) in dialectical materialism are formulated as a series of principles: practice as a criterion for truth, objectivity in considering an object, understanding an object with all its connections, considering an object during its development and motion, and viewing an object as a unity of opposites or dialectical negation, the transition from the abstract to the concrete, the unity of the historical and the logical, analysis and synthesis, and so on [3]. Measurement features have been examined in [4, 5] and elsewhere from the viewpoint of philosophy (from outside). A similar analysis from within is of interest, i.e., from the metrological viewpoint. The final purpose is to formulate initial concepts. Basic concepts in knowledge theory can be specified on the basis of measurement features via the following metrological principles [6]: i. The principle that particular object features are measurable: any quantltyln principle is measurable and under appropriate conditions may be transferred to the class of practically Translated from Izmeritel'naya Tekhnika, No. i, pp. 6-8, January, 1988.
8
0543-1972/88/3101-0008512.50
9 1988 Plenum Publishing Corporation
measurable objects (measurable intensities [7]). This position corresponds to the principle of transferring from the abstract to the concrete. 2. The principle of the unity of the objective and subjective in measurement results means on the one hand that a measured feature reflects a property that really exists and on the other that the concept of a measured quantity is formulated in the consciousness of a researcher and is expressed by certain conceptual facilities. One also assumes unity of the objective and subjective in the content and structure of measurements: all components of a measurement are divided in an obvious fashion into objective (object, means of measurement, experiment conditions) and subjective (related to the human factor), with the latter having a central place. An important consequence is that measurement information is indefinite. 3. The principle that measurement results are conditional means that a measurement resuit is determined by the apparatus and the initial information. 4. The principle that a measurement result is conditionally invariant means that the result provides an adequate estimate of the measured quantity if the method of executing the measurement is properly designed, and if the measurement is repeated, the result should be reproduced with an uncertainty due to all the influencing factors. The result is not dependent on the methods and means of measurement used, which is a concrete expression of the principle that the discussion is objective. However, the independence is restricted to the framework of our existing knowledge on the relation between the object and its environment. Relationships are detected gradually, and their reliability is tested in practice and determined not only by the level of practical achievements but also by the tasks in hand and of course by the perfection of the measurement method. These principles systematize the main gnoseological concepts in theoretical metrology. They involve numerous limitations [2]. The first two principles are the basis for the postulate of [2]: are a definite measurable quantity and the true value of it.
in this model, there
The measured-quantity concept is introduced for a definite model, which is constructed for a particular practical task and reflects features important to that task subject to some property idealization. The model may be a mathematical expression, a scheme, or a nominal description with statement of individual parameters (scalar or vector) corresponding to the particular properties. The model may need to be refined if the measurement task alters. Also, the second and third principles mean that such refinement should be carried out while the object is being examined for new purposes, as the limits to each theoretical concept are relative. Some object features can be estimated more accurately or the measured quantities themselves may be redetermlned. Consequently, different values for a measured quantity may be true for a given object, i.e., one has to recognize that there may be a set of true values for a particular measured quantity. The determinations of a particular measured quantity and the true value are retained while the model is recognized as adequate. A new model does not as a rule mean that the previous is abandoned but simply restricts its sphere of application. There are difficulties of principle in defining how far a model adequately represents an object, which can be identified in attempts to handle practical tasks, and in part in attempts to measure quantities corresponding to the model parameters. An example is provided by Michelson's experiment, which was designed to detect the ether by measuring the velocity of light. The results did not fit the original assumptions on the model parameters, which was the starting point for reconsidering concepts on the object, and thus it sometimes leads to redefinition of a particular physical quantity. In measurement--experlment preparation, there are thus the following necessary stages: model choice, determination (definition) for model parameters, and determination of the measured quantity related to a model parameter. The next measurement feature is due to the result being a number. This is based on the unity of the objective and subjective in measurement results and corresponds to considering an object in its development and motion. The understanding process involves qualitative changes corresponding to the successive complexity increase inpractical tasks for which themeasurement
is intended. Any description by means of a model reflects a particular state. Only when one has determined constant model parameters and established how they correspond to certain values of quantities stable in time and space can one transfer to the measurement experiment. The measurement result can be expressed as a number and be stable only for specified model parameters, which provides scope for checking (repeating) the measurement. These considerations lead to the following postulate [2]: the measured quantity is constant. Any quantity can be taken as constant if its variations are unimportant for the particular task. If a variable physical quantity is a measured one, one should choose the model parameter such as the amplitude and frequency of an alternating current or some functional of the current values such as the effective current. If there is a complicated structure (variability during the examination or use), stable features must be identified for the given property, which can be represented by constant measured quantities. Any model and its parameters represent an approximate description of real object properties; moreover, when one employs the principle that results are conditional and remembers that knowledge is infinite, one can say that an exact model cannot be constructed and that any model is inexact. In measurements, there is an essential constraint on the accuracy with which the model parameters as measured quantities can be stated. This concept is formulated as a postulate [2]: There is a discrepancy between the measured quantity and the object property. Here we note that choosing or making a means of measurement is based on concepts on the measured quantity, i.e., the model, and the measurement itself is made with the facility interacting with the object, namely with the real property. Consequently, the true value for the measured quantity is always determined inexactly, with some uncertainty. In other words, there is no absolutely exact value for the measured quantity (the true value of ultimate instants). Model refinement (if necessary) and measured-quantity redefinition enable one to reduce this discrepancy and thus improve the method and means of measurement. These considerations give a postulate that inexplicitly follows from the first measurement principle: it is always possible to increase measurement accuracy. An important conclusion follows from this. A measured quantity is introduced as amodel parameter, with the model constructed from initial information. The model imperfection is responsible for errors. As the model is determined from a priori information, the attainable measurement accuracy is determined by the data volume on the object and measured quantity. This postulate system may be extended as follows. Firstly, measurement principles imply that measurement is an active procedure performed by man for his purposes. Man inevitably participates in preparing the measurements and interpreting the results. It seems desirable to postulate that there is a relationship of man to the participation in the measurement procedure. This is important in connection with production automation and the creation of engineering areas where human participation is minimal and is of intermediate status. Secondly, eplstomological principles imply that measurement results are objective. This can be so if various conditions are met, including the use of generally accepted measurement units. That condition is of general character and can be proposed in relation to the corresponding postulate. Researches on these lines involve defining the bounds to the measurement concept and the list of obligatory components. In particular, one has to consider the necessity of using means of measurement and the conditions ensuring unified measurements. On the whole, these concepts and their consequences appear adequate to handle the main aspects of theoretical metrology. LITERATURE CITED i. 2. 3.
10
V . M . Sviridenko, Izmer. Tekh., No. 5, 6 (1971). V. A. Granovskii and L. I. Dovbeta, in: Fundamental Metrological Problems [in Russian], NPO "VNIIM im. D. I. Mendeleeva," Leningrad (1981). A. P. Sheptulln (ed.), Dialectical and Historical Materialism [in Russian], Vysshaya Shkola, Moscow (1985).
4. 5. 6. 7.
M. E. Omel'yanovskli, Dialectics in Modern Physics [in Russian], Nauka, Moscow (1983). L. M. Gutner, Philosophical Aspects of Measurement in Modern Physics [in Russian], Leningrad State Univ. (1982). V. A. Granovskii et al., Izmer. Tekh., No. 2, 62 (1986). A. A. Fridman, The World as Space and Time [in Russian], Nauka, Moscow (1965).
MEASUREMENT UNCERTAINTIES AND THE TRUE VALUES AND ERRORS OF PHYSICAL QUANTITIES V. I. Pronenko
UDC 53.081.088.3:389.14
Measurement engineering and metrology are based on the concepts of physical quantity, true values, measurement, errors, and measurement uncertainty. Recently, the specialist literature has increasingly dealt with the need to reconsider these basic concepts, which are recorded in textbooks, monographs, papers on methods, and the standard of [i]. It is suggested [2, 3] that a physical quantity is not a definite property of reality but a parameter in a model. It has also been suggested [2-4] that the true value is not necessary for measurement engineering and metrology because it is unproductive, since it cannot be determined exactly by empirical means etc. It has even been used [3] as an Inexplicit call to God. For one reason or another, it has thus been proposed [2-5] that one should abandon the concept of measurement error and replace it by that of measurement uncertainty. Here, however, measurement uncertainty is not defined clearly. Consequently, basic concepts in measurement engineering and metrology have become debatable. The suggestion that we should reconsider the term physical quantity and assume that it is a parameter in a model adequately representing reality should be considered as the suggestion that we should reconsider the object of knowledge provided by measurement and reconsider the purposes of measurements and the meanings of the results they give. In that context, adequacy is not identity, since a model is a thought product that approximately reflects reality in consciousness, the reflection containing only some major features and their relationships, etc. Undoubtedly, it is impossible to measure a physical quantity without relation to a certain knowledge system expressed as a model. The model parameter obtained during measurement thus represents an intermediate result, which tends to complement the concept that the model is not an adequate representation of reality. Measurement accuracy indices are obtained by considering the accuracy in estimating the model parameter and the accuracy with which the model reflects reality. A model is a prism through which we view reality during measurement, which is an object of recognition. However, the prism is only an instrument, not the object itself. That meaning for a physical quantity agrees with the generally accepted concept that measurement is a cognitive process and corresponds to the generally recognized and legally standardized term [i]. The proposed change in the physical-quantlty concept is impermissible for theoretical metrology because it involves interchanging the object and the purpose of the measurement, and the same applies to measurement engineering and practical metrology, since it rules out the need to incorporate the inadequacy of the recognition and the model in the accuracy parameters. The model is also related to the concept of true value for a physical quantity, and the inadequacy with the uncertainty over the true value. The accepted terms legally defined by the standard of [i] do not include the concept of true-value uncertainty, which has led to incompleteness and incorrectness in defining the term true value for a physical quantity. At present, the true value is defined as the value ideally reflecting the corresponding property of reality. The meaning of the words ideally reflecting is not defined. I consider that the model parameter corresponding to the physical quantity concept reflects reality only approximately and partly in particular measurement tasks, and thus allows various different interpretations, i.e., is characterized by uncertainty. The true value of a physical quantity is to be considered as that value of the corresponding parameter in the Translated from Izmerltel'naya Tekhnika, No. i, pp. 8-9, January, 1988.
0543-1972/88/3101-0011512.50
9 1988 Plenum Publishing Corporation
ii