I. Introduction
Fuzzy Sensor is a term that appeared in the late 1980s. With the development of fuzzy theory technology, fuzzy sensors have also received extensive attention from scholars at home and abroad. Fuzzy sensor is a smart sensor that outputs measurement results in the form of natural language symbol description based on the numerical measurement of classic sensors, through fuzzy inference and knowledge integration. It is generally believed that a fuzzy sensor is a sensor device based on a numerical value that can generate and process symbol information related to its measurement.
2. Research significance of fuzzy sensors
The traditional sensor is a numerical sensor, which maps the measured to a real number set, and uses numerical symbols to describe the measured state, that is, to give a quantitative description of the measured object. This method is accurate and rigorous, and it can also give many quantitative arithmetic expressions. However, with the continuous expansion and deepening of the measurement field, due to the multidimensional nature of the measured object, the complexity of the analyzed problem or the direct acquisition of information, Difficulties in storage and other reasons, only performing simple numerical measurements and describing the measurement results with numerical symbols, this has great disadvantages, such as:
(1) Some information is difficult to describe with numerical symbols. For example, in product quality assessment, people often use "excellent", "sub-optimal", "qualified", and "unqualified". They can also be described by numbers 1, 2, 3, and 4, but the numbers here have lost the usual The meaning of the measured value, it is only used as a symbol and cannot be used to characterize the specific characteristics of the measured entity.
(2) Many numerical measurement results are not easy to understand. For example, when measuring human blood pressure, people are more concerned about whether the blood pressure of the elderly is normal, and whether the blood pressure of the young is high. However, the measured data is often not readable by ordinary people, and therefore cannot meet people's needs.
Therefore, it needs to be supplemented with new measurement theories and methods. Fuzzy sensors are proposed to meet the needs of human life practice, production and scientific practice.
Third, the theoretical basis of fuzzy sensors
1. The principle of symbolic representation
Fuzzy language is a kind of human expression language, because people have a certain degree of ambiguity in the understanding of things in nature. Using fuzzy symbols to express information has the advantages of simple, convenient, and easy to carry out high-level logical reasoning. Fuzzy symbolic representation is the process of using the theories and methods of fuzzy mathematics to describe the measured information with fuzzy language symbols suitable for people's fuzzy concepts with the help of special technical tools. Symbol is the carrier of information, it is the description of the state of an object or event, it defines the characteristic attribute of the entity or the relationship between the entities. Let Q be the numerical domain and S be the language domain. There are several elements qi and si in their respective domains, which are expressed as:
Q = <q1, q2,...> qi∈Q (1)
S = <s1, s2,...> si∈S (2)
At the same time, define a set of relationship families on the domains Q and S:
R=〈R1, R2, …, Rn〉 Ri Q×Q×…×Q (3)
P=〈P1×P2×…×Pn〉 Pi S×S×…×S (4)
And define = 〈Q, R〉, L = 〈S, P〉
Among them, D—Object Relation System, which describes the elements of the numerical domain and their mutual relations;
L—symbol relation system, describing symbol domain elements and their mutual relations.
There are two mappings M and F, M: Q→S, making Si=M(qi), F:R→P, making Pi=F(Ri), and MQ×S and (qi, si) M, It is said that si is a symbol of qi. The meaning of si is the projection of qi from the numerical domain to the language domain, and for every measurement of qi, the symbol si becomes the description of qi.
If the F mapping is a one-to-one mapping and the M mapping is a homomorphic mapping, then there must be an inverse mapping: F-1(Pi)=Ri, M-1(si)=qi. M mapping can be "one-to-one" or "many-to-one" mapping. Then, in the latter case, a symbol in the symbol domain mapped to the numerical domain by M-1 corresponds to not a point, but a "subdomain". Therefore, fuzzy symbolization has certain limitations, that is, under different measurement structures, the elements of the same measurement subset correspond to different symbols; or under the same measurement structure, some elements of the measurement subset correspond to different symbols at the same time Case. This limitation can be compensated by multi-value symbolized measurement based on multi-value logic theory. The basic idea is: in the entity measurement set, according to the degree of performance of a certain feature of the entity, the elements in the measurement subset Q are classified into a certain subset according to the maximum feature membership degree, and the performance of other features is ignored, so As long as you select appropriate multiple feature representations for the entity set on the measurement set to correspond to the elements in the measurement set, you can divide Q into a limited number of subsets {Qi} with related meanings and different performances, and symbolize each Qi Mapping, so as to realize the multi-value symbolized measurement of the entity set.
2. Multi-level mapping principle
Although symbols have the characteristics of high-level logical expression, easy to understand, easy integration of human experience and knowledge, and wider redundancy, compared with the infinitely separable numerical measurement, the degree and scope of the detailed description of the symbolic measurement is not enough, especially in the use of Sign-to-value conversion is more prominent when realizing quantitative measurement. The principle of multi-level mapping can expand the degree of detail and range of symbol representation while realizing the conversion of value-to-symbol and symbol-to-value.
The basic function of level mapping is to realize the transformation of value→symbol and the transformation of sign→numerical value. Its principle is shown in Figure 2. Its information transmission is divided into two situations:
The first is the conversion from value to symbol, and the element qi in the value domain Q is mapped to the subset Si of the symbol domain S through the first level M1 of the mapping M. If the description of the subset Si is not detailed enough, the second step can be performed. Level mapping M2, mapping M2 maps qi to the second subset Sij, after several levels of mapping, the symbol sy describing qi information can be obtained;
Secondly, it is the conversion of the symbol to the value. The symbol sy obtained by the multi-level mapping is mapped to M-1 to obtain the digital value qj.
Due to the limitations of natural language expression concepts, it is recommended that the level of multi-level mapping is 3. For example, for a temperature range of 0°C to 100°C, each level adopts 7 concepts, and when the number of mapping stages is 3, the accuracy reaches 0.3°C. For intermediate measurement results that do not require direct participation of people, the number of multi-level mapping stages can be determined as needed. On the other hand, the number of mapping levels also depends on the number of concepts (elements) contained in each level. If the number of concepts in each level is large, the number of mapping levels required is correspondingly small. If multi-level mapping is applied to fuzzy sensor research including numerical output, the number of mapping levels and the nonlinear error of sensor transformation are related, and the number of mapping levels should be determined by the given measurement uncertainty.
Fourth, the structure and implementation of the fuzzy sensor
1. The structure of the fuzzy sensor
The simplified structure diagram of the fuzzy sensor is shown in Figure 3. It can be seen that the fuzzy sensor is mainly composed of a traditional value measurement unit and a value-symbol conversion unit. The core part is the value-symbol conversion unit. However, the fuzzification and conversion of numerical values ​​into symbols in the numerical value-symbol conversion unit must be carried out under the guidance of experts.
2. The realization method of fuzzy sensor
In summary, the realization of a fuzzy sensor is to find a conversion method between the measured value and the fuzzy language, that is, the fuzzification of the value, to generate the corresponding language concept. The so-called language concept generation is to define a fuzzy language mapping as the fuzzy relationship from the numerical domain to the language domain, so as to map the numerical value in the numerical domain to the symbol domain to realize the function of the fuzzy sensor. The linguistic value here is represented by a fuzzy set, and the fuzzy set is composed of the universe of discourse and membership functions. Therefore, fuzzy language mapping requires the fuzzy membership function in the numerical domain corresponding to the corresponding language concept. How to generate concept is the key to realize fuzzy sensor. There are many ways to realize the function of the fuzzy sensor.
Many foreign scholars have discussed the implementation methods of fuzzy sensors. Here are a few brief introductions:
Introduction to Foulloy algorithm: The essence of fuzzy sensor design is the design of fuzzy transformation algorithm, that is, the selection of reference set and fuzzy quantization. The process is to first obtain the first-level numerical/language transformation strategy in the corresponding measurement field according to the knowledge and experience of the expert or skilled worker, and then apply the fuzzy reasoning method to obtain the corresponding membership function. Foulloy proposed a concept generation method based on semantic relations. First, a universal concept is defined by the meaning of the universe of discourse, called a generic concept, so that it corresponds to the main interval in the domain of the numerical domain, and then new concepts are defined on this basis. , In order to generate other semantic values ​​and their meanings, new concepts are automatically generated internally by the language modifier. Foulloy also proposed a fuzzy state sensor based on a known point set through interpolation. Each learning point uses Delaunay triangulation to construct a fuzzy segmentation on the Cartesian product of the measurement space. The triangulation method is used to establish symbols related to the process state. The vague meaning of.
Benoit E et al. discussed the relationship between the semantics of the symbol and the information to be measured in a specific task environment when using symbolic information. They believed that the fuzzy sensor must be constructed according to the measurement relationship and should be reorganized to adapt to different measurement relationships. And proposed the basic concept as a priori information provided to the sensor, the rest of the concept is automatically generated by the design idea. This method retains the relative semantics between concepts, but it cannot guarantee the consistency with the symbol description of the measurement relationship. Therefore, it is necessary to consider the correction of the measurement relationship by the environment. He proposed a functional method based on qualitative learning and description through compound adjustment. Make corrections. He proposed a new method of constructing fuzzy segmentation based on the linear interpolation of Delaunay multi-dimensional space triangulation to establish a fuzzy sensor with multi-element measurement.
Stipanicer D and others believe that the fuzzy sensor is a kind of intelligent measuring device, which is composed of a simple selection of sensors and reasoners, and converts the measured into a signal suitable for human perception and understanding. Since the knowledge base stores a wealth of expert knowledge and experience, it can measure quite complex phenomena with simple and inexpensive sensors.
Five, the application of fuzzy sensors
At present, fuzzy sensors have been widely used and have entered the homes of ordinary people, such as fuzzy control washing machine cloth quantity detection, water level detection, water turbidity detection, water and rice volume detection in rice cookers, fuzzy mobile phone chargers, etc. In addition, fuzzy distance sensors, fuzzy temperature sensors, fuzzy color sensors, etc. are also the results developed by foreign experts. With the development of science and technology and the fusion of scientific branches, fuzzy sensors have also been applied to systems such as neural networks and pattern recognition.
Six, concluding remarks
The emergence of fuzzy sensors not only broadens the classical measurement discipline, but also makes measurement science take an important step towards human natural language understanding. Although fuzzy sensors have some successful applications in the fields of color and distance, they are only scattered and individual, and far from forming a systematic theoretical system and technical framework. Many key technologies for implementing fuzzy sensors have not yet been completely resolved, and extensive measurement work is still needed. Continuing exploration.
BNC Connector, DC Connector, Coaxial Cable Connector,Coax Connector,DC Power Connector
Chinasky Electronics Co., Ltd. , https://www.chinacctvproducts.com