1
What makes sensor devices and microsystems “intelligent” or “smart”?
Abstract
This chapter demonstrates how the technological progress has resulted in complication of sensors and microelectromechanical systems, formation of the elements of artificial intelligence in them, as well as enhancement of functional possibilities. However, many terms and definitions applied for describing such devices are ambiguous, which complicates communication between specialists, makes the search for required information more difficult, and becomes, in the long run, an obstacle to the development of metrology and measuring techniques. The authors propose a new approach for improving the terminology in this field. It relies on an analogy between biological and technical evolution. Applying this approach, the authors have formed a set of basic interrelated terms in the field under consideration.
Keywords
Intelligent sensor device; Metrological self-check; Sensor; Sensor device; Smart sensor device
1.1. Introduction
The explosive development of computer technologies makes it possible to predict a rapid growth in the number of products and objects that interact and are controlled via the Internet. Enrichment of the relations within the human society, and between the society and the environment, creates a basis for transition to a new stage of technology development, which is characterized by the concept of the “Internet of Things.” The term “thing” is interpreted here in the broadest sense of the word. It can be an object (e.g., some kind of manufacturing) or product. It can be animate or inanimate but should be connected with the digital world through communications and identified in time and space.
Important components of the Internet of Things are the so-called “cyber-physical systems” (CPS) in which various objects are combined by multichannel measurement systems with built-in software.
The CPSs include “smart manufacturing,” “smart building,” “smart energetics,” “smart transportation,” “smart life safety,” “smart health care,” “smart safe city,” and so on. Such systems provide self-organization in the following fields:
At the stage of the Internet of Things, rigid hierarchical information structures are replaced by structures in which an efficiency increase is provided due to the self-organization of links of any information levels with each other.
The transition to the Internet of Things and CPSs is usually called the fourth industrial revolution.
In comparison with traditional measurement systems, CPSs have the following special features (Sapozhnikova and Taymanov, 2015):
- • the number of measuring channels in one CPS may be thousands or many hundreds of thousands of units;
- • measurement channels can include microelectromechanical systems (MEMS) and sensors measuring various quantities, either electrical or nonelectrical; as a rule these channels being spaced far from each other;
- • in the process of CPSs operation, measurands and the number of channels can sufficiently change;
- • in CPSs the “cloud technologies” can be applied;
- • in many cases, measurement information has to be transmitted over great distances;
- • as a rule, an access for metrological maintenance of MEMS and sensors that are components of the CPSs is complicated, and so on.
The analysis of the special features listed above leads to a conclusion that it is economically inefficient to support the CPS channels in the operation state using traditional methods of metrological maintenance.
Calibration of each sensor and MEMS in a year (such is a widely used interval between the procedures of metrological maintenance) or even in four years, would require unacceptably high work load. An alternative is to increase significantly the requirements for reliability (in the first place, metrological reliability) of sensors and MEMS embedded in CPS.
The need for significant improvement of metrological reliability of sensors and MEMS is growing fast also in nuclear energetics, astronautics, and some other fields where an access to embedded measuring instruments is connected with risk or practically impossible.
Analysis of the biological evolution has shown that improvement of “reliability of functioning” of living creatures is due to their complication and development of intelligence. The development of the computer technologies opened new prospects for realization of various measuring instruments applying elements of artificial intelligence.
A strong increase of the number of corresponding theoretical and applied papers can be noticed through the last years (Tarbeyev et al., 2007; Taymanov et al., 2017; Taymanov and Sapozhnikova, 2010b; Sapozhnikova and Taymanov, 2015).
Development of science and the associated technical terminology is an integrated process. As T. Kuhn (1962) has shown, the scientific progress is a series of “scientific revolutions” that correspond to spasmodic changes of concepts (paradigm shifts) concerning further development of sciences.
Such regularity is characteristic not only for the fundamental sciences but also for the applied sciences, and in particular the sciences associated with the development of control systems and measuring instruments, including sensors and MEMS.
The “paradigm shifts” relate to terminology too. In this domain, changes are caused by a desire to define the notions, which are used for new concepts. However, the development of the applied sciences is closely connected with the market: scientists and designers create products for the market, and the market, in its turn, stimulates them and invests in the development of new products. This interaction has an impact on the terminology.
Manufactures and dealers tend to use some “embellishing” attributes for advertising their goods. Without sufficient information about the quality of these goods, customers make their choice based on the name of these goods, which in some cases are not adequate. This may contradict the interests of the society as a whole.
Lack of monosemantic terminology, especially under conditions of an immense technological progress, as well as economy globalization, complicates contacts among specialists, makes the search of required information more difficult, and becomes, in the long run, an obstacle for the development of science and engineering. The ambiguity and insufficient number of terms lead to dishonest competition because the names of goods (among other factors) have an influence on the prices of these goods (Taymanov and Sapozhnikova, 2009a).
In this connection, the need to harmonize the terminology in the field of sensors and MEMS, including those with elements of artificial intelligence, becomes increasingly important. It is believed that many living creatures have intelligence, but crows, dogs, and humans are distinguished by different levels of intelligence. In a similar way, sensors and MEMS can have different levels of intelligence too. It is necessary to find a criterion for determining the level of intelligence of these devices, which could be applied for forming a corresponding terminology group.
One would think that а new concept can be defined with the help of a known term supplemented with a set of qualifying adjectives. However, for popular new concepts described with more than two additional adjectives, a new term will inevitably appear. The fact is that in scientific and technical papers, books and particularly in advertisement booklets, part of the supplemental attributes will be voluntarily or involuntarily omitted. Depending on what specific words have been excluded from the text and what the experience of the reader is, the text interpretation can appear to be different. The situation is redoubled because some terms, which were defined many years ago because of the development of technology, have lost their unambiguity (Taymanov and Sapozhnikova, 2009b). The speed of updating terminology vocabularies, including the International vocabulary of metrology (VIM, 2012), as well as terms given in prescriptive documents, remains behind the pace with which new terms appear. Names for new concepts are born and spread in numerous scientific publications. As a result, quite often it is possible to come across terms, which are differently interpreted. On the other hand, in some cases different terms are used for similar concepts.
1.2. Interpretation of terms related to sensors
In recent years, the intensive development of metrology and measuring techniques has resulted in implementing sensors with new functions, a set of which can be significantly different for various types of sensors. And here a question arises with respect to their names, whether they adequately reflect the features of special sensors or not.
1.2.1. About the term “sensor”
According to IEEE (1999), “sensor” means a “component providing a useful output in response to a physical, chemical, or biological phenomenon. This component may already have some signal conditioning associated with it.” (In the same document an explanation is given that the signal conditioning “involves operations such as amplification, compensation, filtering, and normalization”).
“Examples: platinum resistance temperature detector, humidity sensor with voltage output, light sensor with frequency output, pH probe, and piezoresistive bridge.” It should be noticed that it follows from the second sentence of the above definition that two versions are possible: a sensor may have an internal signal conditioning or have no conditioning at all.
In the IEEE document (IEEE, 2003) published later, the term “sensor” means “a transducer that converts a physical, biological, or chemical parameter into an electrical signal.”
Note: According to IEEE (1999) “transducer” means “a device converting energy from one domain into another, calibrated to minimize the errors in the conversion process…”
In VIM (2012) “measuring transducer” is considered to be a “device, used in measurement, that provides an output quantity having a specified relation to the input quantity.”
According to NAMUR (2005) and VDI/VDE (2006), “sensor” is a “device that picks up physical measurement variables and converts them to standardized output signals.”
In VIM (2008, 2012), the term “sensor” is defined as an “element of a measuring system that is directly affected by a phenomenon, body, or substance carrying a quantity to be measured,” i.e., it is a primary measuring transducer. Some elementary examples give an explanation of this term: “sensing coil of a platinum resistance thermometer, rotor of a turbine flow meter, Bourdon tube of a pressure gauge, etc.” In clause 3.8 of VIM (2008), it is stated that in some fields the term “sensor” is replaced with the term “detector,” which, according to clause 3.9, has another meaning.
In the recommendations (RMG 29-2013, 2014), the term “sensor” is considered to be a synonym of the terms “sensing element” and “primary measuring transducer.” Its definition corresponds to the VIM definition. In the Note, it is emphasized that “a primary transducer or combination of a primary transducer and other measuring transducers made as an autonomous assembly is called datchik.” The latter word is translated as “giver” and means that this device gives information.
The ambiguity of the term “sensor” becomes stronger because of a vague definition of the “measuring system” concept.
In essence, in most of the above-mentioned documents, as well as in numerous dictionaries, with which it is possible to become acquainted through the website (Anon, 2017), the term “sensor” can be interpreted as an isolated device or a part of a more complicated device. Exceptions are the documents (NAMUR, 2005; VDI/VDE, 2006).
The ambiguity of this term is aggravated because of development of sensor arrays, such as structures forming artificial finger (Neumann et al., 2015), tongue, nose (Banerjee et al., 2016; Borras et al., 2015), etc.
In practice the term “sensor” is applied for designating (Taymanov and Sapozhnikova, 2009b):
- 1. one sensitive element;
- 2. primary measuring transducer that can include a sensitive element or a group of sensitive elements (e.g., an array);
- 3. measuring transducer that consists of a number of separate transducers, connected in series, e.g., a primary transducer and amplifier;
- 4. isolated unit that can contain any component according to items (1–3) or a group of the corresponding components in any combination;
- 5. unit, according to item (4), which contains also additional signal processing units, e.g., analog-to-digital converter (ADC), bus interface, microcontroller, and indicator (in any combination).
Examples of sensors according to clauses (4 and 5) are sensors-on-chip consisting of sensitive elements and a set of measuring transducers applied in smartphones (Yurish, 2013). They can be built-in (embedded, internal) sensors or external ones and can measure, e.g., blood glucose, body temperature, one-lead ECG, heart rate, blood oxygen saturation, body fat percentage, and stress levels.
Thus, the term “sensor” remains vague. Therefore, in some publications the explanation of specific terms is given. For example, at the beginning of (Duta and Henry, 2005), it is written that “for the aim of this paper a sensor is the device consisting of one or more transducers and a transmitter that converts the transducer signals into a form recognizable by the control or monitoring system.”
In some publications (Abbate et al., 2012; Gundlach, 2007; Spencer, 2012, and many others), to define a structurally complicated device containing at least one sensor and additional transducers, the term “sensor device” is applied.
It is necessary to distinguish the concepts applied for definitions of a sensitive element (sensor) and autonomous device including one or several sensors and other components.
1.2.2. Definitions of key terms related to devices with elements of artificial intelligence
In scientific and technical publications there is a great variety of terms used for the devices with artificial intelligence (Taymanov and Sapozhnikova, 2009a, 2010a), e.g., “intelligent,” “smart,” “adaptive,” “self-validating,” “system-generating,” “self-calibrating,” “self-diagnosing,” “fault-tolerant,” “self-checking,” “cogent,” “brilliant,” and so on (Allgood and Manges, 2001; Barberree, 2003; Bialas, 2010; Denton, 2000; Duta and Henry, 2005; Falconi et al., 2007; Gaura and Newman, 2004; Hans and Ricken, 2007; Henry and Clarke, 1993; Hashemian, 2005; Huijising et al., 1994; Huijsing, 2008; Hunter et al., 2010; Itskovich, 2002; Jackson, 2004; Kirianaki et al., 2002; Langari and Morris, 2012; Makinwa, 2014; Romanov et al., 1994; Wang et al., 2005, Werthschutzky and Muller, 2007; Yurish, 2008, 2010, 2011; Zook et al., 2006; Zvetkov, 1999; and many others). However, none of them was included into the International vocabulary of metrology (VIM, 2012).
In industry, the most widespread attributions for devices with elements of artificial intelligence are “smart” and “intelligent.”
A number of definitions of such instruments were stated in some national standards, recommendations, and international documents, which do not refer to the sphere of metrology (e.g., BSI, 2005; IEEE, 1997, 1999, 2003; NAMUR, 2005; VDI/VDE, 2005). In particular, in accordance with IEEE, (1997, 1999, 2003), “smart sensor” is a “sensor version of a smart transducer.” “Smart transducer” is a “transducer that provides functions beyond those necessary for generating a correct representation of a sensed or controlled physical quantity. This functionality typically simplifies the integration of the transducer into applications in a networked environment” (IEEE, 1997, 1999, 2003). It is unclear from this definition what additional functions are implied. The practice shows that the additional (“redundant”) functions in question can be diverse. The importance of these functions is not the same for various customers, and the number of the “redundant” functions in a given device can be different. Therefore, at present such a definition of the term “smart sensor” is not satisfactory.
In a number of publications, operations fulfilled in the past by a human operator are considered to be among the “redundant” functions of such a smart sensor (e.g., Itskovich, 2002), in particular, zero adjustment, linearization, choosing a necessary measurement range, correcting for an influencing factor using a special sensor, and so on. According to Romanov (Romanov et al., 1994), such devices should be considered as “adaptive.” Specific features of “sensors” and “sensor systems” with self-adaptation capabilities are discussed in Yurish (2008, 2011).
According to Allgood and Manges (2001), “smart sensor” is a programmable “measurement system that has sufficient computational capacity to support the data acquisition, memory, and decision-making necessary to respond to algorithmic instructions.” It responds to calibration requests, reacts “to interrogations about its health and status, and provides uncertainty estimations.” In (Zook et al., 2006) the term “smart sensor” implies that “some degree of signal conditioning is included in the same package as the sensor. On the more sophisticated end of the spectrum, the ‘sensor’ unit can include devices with elaborated signal processing, displays, and diagnostic or self-calibration features.”
In accordance with Kirianaki et al. (2002), “smart sensor is one chip, without external components, including the sensing, interfacing, signal processing and intelligence (self-testing, self-identification, self-validation or self-adaptation) functions.” However, this definition does not distinguish instruments with significantly different consumer features and complexity. For example, it levels
According to Duta and Henry (2005), “intelligent sensor” can “carry out internal diagnostics.” It produces an estimation of measurement value, measurement uncertainty, and “a discrete valued flag indicating how the measurement value and uncertainty have been calculated.” In essence, a similar definition is used inBarberree (2003), where such a device is called self-diagnosing, self-validating, or self-calibrating. In analogous way, Wang et al. (2005) consider the intelligent of those instruments, which are based on fieldbus and can perform such functions as self-calibration, self-compensation, self-validation, etc.
Hunter et al. (2010) discuss features of a “smart sensor system” that may include multiple sensors and note the intelligent features, which can be performed “at the sensor level including but not limited to: self-calibration, self-health assessment, self-healing, and compensated measurements (auto zero, calibration, temperature, pressure, relative humidity correction).” “The ideal goal is to have a self-contained smart sensor system that is cost-effective, reliable, self-monitoring, reconfigurable, and can operate indefinitely.”
An alternative is to define “smart” and/or “intelligent” sensors through a description of their structure. In Di Lecce and Calabrese (2012) this approach is called “hardware-centered.”
According to Makinwa (2014) “smart sensor is a system-in package in which a sensor and dedicated electronics are realized. It may consist of a single chip…. However, in cases when the sensor cannot be implemented in the same technology as the interface electronics, a two-chip solution is required.”
In some publications, e.g., Huijsing (2008), “smart sensor” is considered to be a combination of “a sensor, an analog interface circuit, an analog to digital converter (ADC), and a bus interface in one housing.” In the same publication Huijsing also considers an “integrated smart sensor.” It should contain “one or more sensors, amplifiers, a chopper and multiplexers, an AD converter, buffers, a bus interface, addresses, and control and power management.” Most authors deem “intelligent” or “smart” those devices, which are equipped with a microprocessor or microcontroller (http://www.answers.com/topic/smart-sensor; http://computer.yourdictionary.com/smart-sensor; Bialas, 2010; Falconi et al., 2007; Gervais-Ducouret, 2012; Itskovich, 2002; Jackson, 2004; and many others). Such devices, as a rule, can carry out signal conditioning, filtering of noise, compensating certain error components, producing an alert signal if a measured quantity value reaches a critical limit, etc.
A parallel idea is presented in BSI (2005), where intelligent instrument is considered to be a “device where embedded computing capability is included as part of a measurement, control or actuation device.”
In accordance with Gervais-Ducouret (2011), as opposed to any “smart” sensor, an “intelligent” sensor has a small microcontroller, a memory (flash, RAM, ROM), and optimized architecture for sensor applications. This intelligent sensor concept includes combination of possibilities to process sensor data, to reconfigure embedded functions, and aggregate external sensor data. The sensor, to which additional external sensors are connected for aggregating and joint preprocessing the data, is called “intelligent sensor hub” (Gervais-Ducouret, 2012; Di Lecce and Calabrese, 2012). The intelligent sensor hub usually includes additional analog or digital interface.
Yurish (2010, 2011) emphasizes that the attribute “smart,” to a greater extent, can be related to “technological” aspects of the devices, whereas “intelligent” to “functional” ones. He supposes that in the common case, a smart sensor may have no intelligent functions, as well as an intelligent sensor may not be smart.
Meyer et al. (2009) and Kiritsis (2011) considering intelligent products, emphasize that the terms “intelligent” and “smart” can be applied as interchangeable.
In the book (Langari and Morris, 2012) it is written that “there are no hard distinctions between the function of…intelligent instrument, intelligent sensor, smart sensor, and smart transmitter…and which term is used to refer to an intelligent device is largely due to the preference adopted by different manufacturers for one name or another.”
The terms and their definitions related to MEMS are also various (Enoksson et al., 2005; Gaura and Newman, 2004; Madhavi et al., 2011; Movassaghi et al., 2014; Newman et al., 2006; Shkel, 2001; and others). In industrial applications, the most widely spread attributes for the devices with artificial intelligence are also “smart” and “intelligent.”
An analysis of the “intelligent” and “smart” concepts in a number of widely used English, American, and Russian dictionaries testifies that sometimes both terms are applied in the same sense, but generally the word “intelligent” relates to a higher level of development, than “smart.”
In 2009 a mini survey “What does it mean ‘smart sensor’?” was set up at the Sensors Web Portal. They got a feedback from 227 participants from both academia and industry (Yurish, 2010, 2011). The survey results have shown that 18% of the participants named “smart” those “sensors,” which include combination of sensing element, analog interface circuit, ADC, and bus interface. At least, 67% of experts considered the main features of “smart sensors” to be some combinations of self-identification, self-validation, self-testing, self-adaptation, and/or self-calibration. The survey results allow the main features characterizing the tendencies in sensor development to be revealed. Part of the participants emphasized the extended network capabilities of “smart sensor.”
It should be notified that the survey concerned only the term “smart,” whereas the term “intelligent” was not touched. The answers of the survey respondents could not show the difference in interpretation of these terms, as they were treated as synonyms.
The terminology ambiguity often brings customers to confusion in their attempt to make the best choice. Moreover, this ambiguity can become an obstacle to develop new equipment. The terms are blurred acquiring polysemy because new challenges emerge, with which these terms are semantically associated. Therefore, the most important matter in designing a concept of an international vocabulary of terms is to take into consideration trends of developing measurement techniques and to reveal the problems, which will determine the vocabulary of specialists in the field of measurements and control over many years.
1.3. Key trends in the development of sensors (sensor devices) and microelectromechanical systems
What is the driving force of sensor and MEMS development? Why are they called “smart” or “intelligent” more and more often?
1.3.1. The method of analogy
To answer these questions let the analogy be applied between biological organisms and technical means, as well as between the processes of their evolution. The suggested methodological tool seems to be helpful because the technical means are on a significantly lower stage of their development than the most perfect biological organisms.
In their books, Wiener (1965) and particularly Lem (1980) who was not only a writer but also an outstanding philosopher demonstrated the efficiency of applying this analogy (Taymanov and Sapozhnikova, 2003, 2009а). At present, it is very successfully used in evolutionary cybernetics (Red'ko, 2007; Turchin, 1977), as well as in bionics (biomimetics), which allows functions and structural elements resembling solutions made by nature earlier, to be realized in technical means (devices) (Bogue, 2009; Nagel, 2016; Stroble et al., 2009; Nagel et al., 2010).
The similarity between biological and technical evolution, forms not only a “reference book” containing successful structural and functional decisions but it also gives a necessary strategic criterion for development of engineering (Taymanov and Sapozhnikova, 2008). Prof. Popper (1984) has noticed that all organisms are “problem solvers”; the problems are born together with beginning of life on earth. Evolution selects the best solutions of the problems.
1.3.2. Complication of organisms and sensors as a tendency of evolution
The key purpose of each living organism is to prolong its own life and existence of the species. As the duration of life grows and geographic range widens, the amount of possible dangerous situations and their variety increase. Accordingly, in the process of evolution, the methods selected for reacting to dangerous situations also become enriched. In nature, an “evolutionary change is not a continuous thing; rather it occurs in fits and starts, and it is not progressive or directional” (Ward, 2001). For example, organisms got smaller, as well as larger, “materials” of a capsule, shell, or coat changed. On the other hand, evolution “has indeed shown at least one vector: toward increasing complexity” (Ward, 2001). Primarily, conservative methods originated, later adaptive methods arose, and, at last, methods based on intelligence came into being.
This vector is also traced in the development of biological sensors from the simplest ones, which perceive touching, to sensor arrays, from a sense of touch to a wide spectrum of perception, and then to multichannel sensor systems with various sensors, which allow more complicated parameters (e.g., color, odor, etc.), including those distributed in space, to be estimated.
It should be noticed, that the movement along this vector happened with ever-increasing average speed.
From an evolutionary point of view, there is no barrier between animals and humans, when methods for ensuring the survival are considered. “Perhaps that is the fate of future human evolution: greater complexity through some combination of anatomy, physiology or behavior” (Ward, 2001).
Such a point of view corresponds to the position shared by Klix (1980), McFarland (1999), Hodos (1982), Asimov (1994), as well as many other biologists, philosophers, and psychologists. According to this position, the intelligence is a means providing the most developed abilities of organism survival, including adaptation to current and expected events (adaptation to climatic changes and intrinsic changes caused by aging, etc.). Therefore, intelligence enables a life span to be extended. Over the last several hundred years the human life span has doubled. It is logical to associate such an extension with the increasing role of intelligence in the human life, as well as with the knowledge obtained using it.
In the process of evolution, it was just intelligence that appeared to be the most powerful factor contributing to the survival of living creatures under changing environment.
The similarity between the biological and technical evolution is confirmed even in their details. The block-module principle of organization and development of molecular genetic systems shows that new systems are formed “bottom-up” with macromolecular components playing the role of building blocks/modules (Red'ko, 2007).
Similar processes took place in the development of sensor devices with the capabilities to
- • interact with other devices,
- • to adapt to a changing environment,
- to perform a self-check (in particular, metrological self-check for measuring instruments), and forecast their health on the basis of artificial intelligence. In sensor devices, separate sensing elements, additional components, ADCs, microcontrollers, and so on are combined into a single entity.
By developing the idea of the analogy between the biological and technical evolution, it is possible to consider a direct analogy between the life span of living organisms and the lifetime of sensors and MEMS, during which sensors and MEMS are characterized by their good operating mode and the absence of any maintenance requirement.
Over a long period of operation and in a wider field of application, various potential defects, which have not been noticed during the manufacturing process, and deviations of device characteristics may reveal themselves. Such defects may be caused, for example, by wear or unforeseen changes of external factors. In sensors and MEMS applied for performing measurements, such defects and parameter deviations lead to increase of the uncertainty and degradation of the reliability of measurement results.
The artificial intelligence (at a certain level) should neutralize actively the above faults; contribute to widening the range of operating conditions; and extend the lifetime of the sensors and MEMS without any maintenance. Eventually it should enhance the reliability of the measurements and increase the calibration interval.
The idea of applying the elements of artificial intelligence in sensors and MEMS, which are, as a rule, the weakest points of any equipment, appeared not long ago and started developing on the basis of modern technological achievements. At this stage, it became possible to increase the structural complexity of these components and to extend considerably the lifetime of the instruments and systems, which incorporate them (Taymanov and Sapozhnikova, 2010b; Sapozhnikova et al., 2016).
Table 1.1 shows methods for providing the reliability of functioning in the process of biological evolution (as applied to fauna) and technical evolution (as applied to sensors).
Certainly, new materials and technologies will be developed in the future. They will allow the realization of more stable device characteristics. Undoubtedly, the reliability requirements of sensors and MEMS will increase, as well. Both these factors will result in keeping the trend of development of their structures (from more simple structures toward more complicated ones). It should be emphasized that in general the “complexity” concept has to take into account both structural and technological aspects. For example, to obtain some useful effect in a sensor, the application of a film with a spatially homogeneous structure (practically, without any defects), or a piece of extra pure metal, can mean a sensor complication.
1.3.3. Features and forms of intelligence
As pointed out above, intelligence is an efficient means to prolong the existence of an intelligence “carrier.” Intelligence implies an ability to perceive and transmit information, accumulate and analyze it, as well as make decisions.
The first feature of intelligence is the capability of its “carrier” to interact with the surrounding environment, including other members of population, on the basis of the results of processed input information. For sensors and MEMS with artificial intelligence, applied in modern measuring and control systems, this feature characterizes the capability:
The second feature is the ability of the intelligence “carrier” to change his/her characteristics under the influence of the obtained information. For the considered devices with artificial intelligence, this feature requires a possibility to change parameters and/or the operation algorithm in accordance with the signals obtained from embedded sensors or external sources.
The third feature is a capability of the intelligence “carrier” to estimate his/her own health and to make decisions improving the chance for survival. For sensors and MEMS with artificial intelligence this feature corresponds to a self-check ability, including metrological self-check ability. In many cases the results of the metrological self-check can be used for performing self-correction and self-recovery, as well as estimation and prediction of the metrological health.
It is noteworthy mentioning that the above three features can be related to both humans and devices with the elements of artificial intelligence. Accumulation of the above features demonstrates an increase of the level of natural or artificial intelligence.
The noted features remind to a certain extent the levels of product intelligence according to Kiritsis (2011), but this paper does not consider the analogy of products and living creatures.
Table 1.1
In our view, in the hierarchy of these features the third one is the most important for customers. In essence, it is a question of an ability to check up the reliability of information obtained from sensors.
In nature, intelligence has developed itself in two ways: formation of a “collective mind” of many living organisms and development of intelligence of a separate individual. If the risk of extinction of individual living organisms is high, the “collective mind” provides a way of preserving the experience gained and supporting the life of the species as a whole. Certainly, this is the type of intelligence, which originated at the early stages of the evolution, when the life span of an individual organism was short. A typical example is given by swarming insects, e.g., bees, which select the reliable information by “voting.” The importance and validity of the information obtained by scout bees depends on the number of bees obtaining this information (McFarland, 1999). By the way, the origin of “democratic methods” in the human society for estimating the reliability of vitally important information and subsequent development of this method for controlling a state demonstrate the efficiency of applying the experience from nature (Taymanov and Sapozhnikova, 2008).
The formation of the “collective mind” is a typical example of a metasystem conversion representing the integration of a number of autonomous subsystems at a lower level (they can be different, to some extent) and development of an additional control mechanism at a higher level (Red'ko, 2007). A similar approach is frequently used in the metrological practice for checking the reliability of information. For example, in nuclear power plants, a redundant number of sensors are joined into a multichannel system (Hashemian, 2006). The metrological “health” of a sensor can be estimated on the basis of evaluation of its signal deviation from the signals of the remaining sensors included into the system. Information from the majority of the sensors is considered to be valid. The “swarm” model is applied, e.g., in geology and other fields, to reveal reliable information (Bogue, 2009).
However, selection of reliable information on the basis of the “swarm” model has turned out to be not the most effective one in the biological evolution due to insufficient flexibility under changing environment (Taymanov and Sapozhnikova, 2008).
The development of intelligence of an individual living creature was accompanied by an increase in its life span. Intelligence has been formed within the framework of a metasystem development as a result of integration of a number of independent nervous structures, each of them being linked with one or several biological sensors and distinguishing themselves by individual features. The “health” of these sensors is checked with the help of intelligence. In comparison with the “collective mind,” the intelligence of an individual has an advantage in searching for effective ways to survive under a changing environment.
The brain has a special mechanism for testing the stability of essential activity characteristics. This mechanism, known as an “error detector,” has been discovered by the famous Russian Academician Bechtereva et al. (2005). A person diagnoses (feels) a “malfunction” of his organism, including a “malfunction” of a sense organ, such as an eye or ear, initially, through an unpleasant sensation caused by signals coming from sensors or indirectly from the “error detector.” After that he/she makes a corresponding decision, e.g., to irrigate an eye or take a medicine. In fact, the feeling of an indisposition characterizes an interaction of the brain with individual organs and gives the basis for making the decision on treatment.
As applied to technical device, this interaction corresponds to the definition of intellectuality as an ability, based on knowledge, to analyze input information and prepare the basis for a decision regarding future actions (Andrew, 1983; Romanov et al., 1994).
Besides the method discussed above, a human, as well as other developed living creatures, applies auxiliary ways for detecting deterioration in operation of the sensing organs, such as:
Thus, in the biological evolution, both types of intelligence considered above are developing themselves in parallel. Sometimes, they supplement each other, but intelligence of an individual has gained the priority and obtained a greater pace of improvement. In the human society, the “collective mind” plays a stabilizing role. The same role it plays in multichannel systems with artificial intelligence. The intelligence of individual provides a greater ability to survive for a particular individual organism and contributes to survival of species as a whole.
The similarity of the biological and technical evolution gives the basis for forecasting future trends in sensor devices and MEMS. For example, in the nearest future one should expect a development of self-studying intelligent devices with the self-check of “health” (in the first place, metrological “health”). Such instruments will be able not only to correct consequences of “aging” and influencing factors but also to adjust their parameter values automatically on the basis of predicted changes. This can further enhance the efficiency of their operation. In technical complexes, sensor devices with the self-check, combined into a system with a “collective mind,” will be widely used. Soon sensor arrays susceptible to one and the same quantity, as well as arrays combining groups of sensors susceptible to various quantities, will find a wider application (Taymanov and Sapozhnikova, 2008, 2009b).
1.4. Suggestions for improving terminology in the field of sensors and microelectromechanical systems
The concept of history and trends of sensor and MEMS development, based on the analogy between biological and technical evolution, gives grounds to form a systematized terminology to overcome or, at least, to reduce the existing ambiguity.
Below a proposed set of the most important terms and their definitions related to the subject of the present book are given with short comments.
“Sensor”: an element (a primary measuring transducer) of a measuring or control instrument, which is directly affected by a phenomenon, body, or substance carrying a quantity to be measured. It is the simplest sensitive element. (The definition is close to the definition given in the International vocabulary of metrology (VIM, 2012) and corresponds to item (1) of the list given in Section 1.2.1).
The definition of the term “measuring transducer” applied hereinafter corresponds to VIM (2012) and is given also in Section 1.2.1.
“Sensor device”: a self-contained device including one or more sensors, and in some cases additional components (e.g., an amplifier, filter, ADC, interface, etc.). This definition is close to the definition of the Russian term “datchik.”
“Measuring system”: a group of sensor devices and other components associated with them, having a common data processing unit, for measuring one or more quantities in various points of space. This definition corresponds with the meaning given in RMG 29-2013 (2014).
“Metrological self-check of a sensor device”: an automatic procedure of testing whether the uncertainty or error of a sensor device in the process of its operation lies within some permissible limits specified under operating conditions. This procedure is carried out using a reference value generated with the help of an additional (redundant) embedded component (a sensor, secondary transducer, or material measure) or additional parameter of an output signal.
The term “reference value” corresponds to the same term given in VIM (2012). The metrological self-check can be realized in two forms: a direct or diagnostic self-check (Taymanov and Sapozhnikova, 2005; Taymanov et al., 2011). If the metrological self-check is accompanied by evaluation of uncertainty or error, it is usually called “self-validation” (Sapozhnikova et al., 2005).
“Smart sensor device”: a sensor device capable to interact with other devices on the basis of processing the results of information received. Such a definition corresponds to the first (simplest) feature of intelligence given in Section 1.3.3 of this chapter. Here the concept “interaction” implies a possibility of noise-immune communication with other remote devices. In its turn, the noise-immune communication, as a rule, is realized by coding information (an analog signal is converted into a digital form in the sensor device). This provides the information reliability when transmitting it over longer distances. An example can be a sensor device containing a sensor, ADC, interface for communication with a common bus, and some other components.
“Smart sensor hub”: “smart sensor device,” to which additional external sensors are connected for joint data processing. This instrument can be considered as a version of a measuring system.
“Adaptive sensor device”: smart sensor device, the parameters and/or operative algorithms of which can change in the process of operation, depending on internal signals of additional sensors and secondary transducers, which this device contains, and/or on external signals.
An adaptive sensor device can provide its adaptation (adjustment) to a measurement range, speed of measurand variation, effect of influencing quantities, etc. within the limits specified in a technical documentation. The definition of “adaptive” sensor device supposes realization of the second feature of intelligence.
All the capabilities listed above provide working capacity of an adaptive sensor device within an expanded range of environmental conditions. Adaptation to changing conditions enables correcting output signals according to the relationship known in advance. Thereby this adaptation contributes to extending the lifetime of sensor devices and MEMS without any maintenance. The reliability of measurement results obtained from adaptive sensor devices applied as measuring instruments increases, and the calibration interval can be extended.
“Intelligent sensor device”: smart sensor device with the function of a metrological self-check.
At present, such a device is an instrument that can enable the highest level of information reliability. According to the above classification, it can be characterized by all three features.
Additionally, in the intelligent sensor device, it is possible to realize (Taymanov et al., 2011):
- • forecasting the metrological “health” of the sensor device taking into account the results of the self-check for a previous time interval;
- • automatic correcting an uncertainty or error caused by influencing quantities, the value of which is unknown, and/or aging of some device components (in a number of cases);
- • extending the calibration interval on the basis of the results of self-checking, forecasting the metrological “health” and automatic correcting;
- • performing self-recovery of the sensor device if a single defect appears in it (in a number of cases).
“Self-recovery”: automatic procedure for damping consequences of an occurred defect, i.e. it is a procedure of providing a fault-tolerance.
“Fault-tolerance”: ability of the sensor device to keep specified characteristics within permissible limits when a single defect has occurred.
Methods of the metrological self-check and examples of intelligent sensor devices are considered, in particular, in Taymanov et al. (2011). Sometimes a need arises to separate in space a sensor device from a data processing unit. Such a need can be caused by a harsh operating environment of the sensor device, which is unacceptable for microprocessors. The same situation can also arise if a centralized data acquisition in a measuring system is required.
Intelligent sensor devices provide the longest lifetime without maintenance and “healthy” state within a wider range of environmental conditions.
Unfortunately, the semantic difference between the terms “intelligent” and “smart” when they are translated into a number of widespread languages, e.g., German and Russian, is significantly less than the difference in customer characteristics of the corresponding sensor devices. This fact can lead to certain difficulties.
The main features of the definitions given above can be applied to a measuring system, particular forms of which can be smart, adaptive, and intelligent system on the basis of sensor devices and MEMS.
Corresponding proposals regarding the terms applied in metrology were prepared for a future edition of the International vocabulary of metrology. A part of the terms and definitions given above was included into the Russian standard (GOST R 8.673, 2009).
Industrial production of sensors and MEMS with artificial intelligence does not keep pace with development of new devices.
Smart sensors have been produced for many years in large quantities. As for adaptive sensor devices, their variety is not too wide and they are met not so often. At present, intelligent sensor devices are mostly at the stage of development.
Fig. 1.1 illustrates the increase of interest, the specialists show particularly in the development of intelligent sensor devices corresponding to the above definition. The statistical data were obtained using the bibliographic database “Scopus.” The vertical axis presents the percentage of publications in the field of “measurement,” which contain the term “metrological self-check” or words close to it in meaning. Medicine, neuroscience, materials, psychology, economics, veterinary science, and some other fields were excluded from the analysis.
1.5. Conclusion
Sensors and MEMS with the elements of artificial intelligence permit operational costs to be reduced but cause a number of new problems. To solve them it is necessary to provide specialists with the possibility to share their experience, to define the performance requirements and characteristics of new devices, as well as to develop new normative documents (standards, guides, and recommendations).
The above analysis of using the terms applied for sensors and MEMS with the elements of artificial intelligence shows that the ambiguity of terminology impedes the solution of these tasks. In this connection harmonization of the terminology is of significant importance.
Arguments for applying the analogy between the biological and technical evolution are given to reveal the tendencies of sensor development. A set of interrelated terms and their definitions, important for practice, is proposed on the basis of this analogy, taking into consideration the results of terminological analysis.
Acknowledgment
The authors of the chapter express their deep gratitude to Dr. Nihtianov for the useful and interesting discussions.
..................Content has been hidden....................
You can't read the all page of ebook, please click here login for view all page.