Abstract
Fuzzy methodology transforms expert ideas—formulated in terms of words from natural language—into precise rules and formulas. In this paper, we show that by applying this methodology to intuitive physical and mathematical ideas, we can get known fundamental physical equations and known mathematical techniques for solving these equations. This fact makes us confident that in the future, fuzzy techniques will help physicists and mathematicians to transform their imprecise ideas into new physical equations and new techniques for solving these equations.
Keywords
These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.
This is a preview of subscription content, log in via an institution.
Buying options
Tax calculation will be finalised at checkout
Purchases are for personal use only
Learn about institutional subscriptionsReferences
Aczel, J.: Lectures on Functional Equations and Their Applications. Dover Publication, New York (2006)
Akhiezer, A.I., Berestetsky, V.B.: Quantum Electrodynamics. Interscience Publishers, New York (1965)
del Catillo-Mussot, M., Costa Dias, R.: Fuzzy sets and physics. Revista Mexicana de Física 39(2), 295–303 (1993)
Feynman, R., Leighton, R., Sands, M.: The Feynman Lectures on Physics. Addison Wesley, Boston (2005)
Gutierrez, E., Kreinovich, V.: Fundamental physical equations can be derived by applying fuzzy methodology to informal physical ideas. In: Proceedings of the 30th Annual Conference of the North American Fuzzy Information Processing Society NAFIPS’2011, El Paso, Texas, 18–20 March 2011
Klir, G., Yuan, B.: Fuzzy Sets and Fuzzy Logic: Theory and Applications. Prentice Hall, Upper Saddle River, New Jersy (1995)
Kreinovich, V., Chang, C.-C., Reznik, L., Solopchenko, G.N.: Inverse problems: fuzzy representation of uncertainty generates a regularization. In: Proceedings of the 1992 Conference of the North American Fuzzy Information Processing Society NAFIPS’92, Puerto Vallarta, Mexico, 15–17 December 1992, NASA Johnson Space Center, Houston, TX, pp. 418–426 (1992)
Kreinovich, V., Quintana, C., Reznik, L.: Gaussian membership functions are most adequate in representing uncertainty in measurements. In: Proceedings of the 1992 Conference of the North American Fuzzy Information Processing Society NAFIPS’92, Puerto Vallarta, Mexico, 15–17 December 1992, NASA Johnson Space Center, Houston, TX, pp. 618–625 (1992)
Nguyen, H.T., Kreinovich, V., Bouchon-Meunier, B.: Soft computing explains heuristic numerical methods in data processing and in logic programming. In: Medsker, L. (ed.) Frontiers in Soft Computing and Decision Systems, pp. 30–35. AAAI Press, Palo Alto (1997). (Publication No. FS-97-04)
Nguyen, H.T., Walker, E.A.: First Course on Fuzzy Logic. CRC Press, Boca Raton (2006)
Ross, T.J.: Fuzzy Logic with Engineering Applications. Wiley, New York (2010)
Zadeh, L.A.: Fuzzy sets. Inf. Control 8, 338–353 (1965)
Acknowledgments
This work was supported in part by the National Science Foundation grants HRD-0734825, HRD-124212, and DUE-0926721.
Author information
Authors and Affiliations
Corresponding author
Editor information
Editors and Affiliations
Appendices
Appendix
Appendix 1: Variational Equations
General derivation. Let us recall how we can transform the Least Action Principle into a differential equation. Let us first do it on the example on Newton-type situation, where we need to find a function \(x(t)\) that minimizes the following expression: \(S=\int L(x,\dot{x})\,dt\rightarrow \min \). Minimizing means, in particular, that if we take any function \(\varDelta x(t)\) and consider a function \(S(\alpha )=x+\alpha \cdot \varDelta x\), then this function must attain its maximum for \(\alpha =0\). Thus, the derivative of \(S(\alpha )\) at \(\alpha =0\) must be 0. Differentiating the expression
and equating the derivative to 0, we conclude that
Integrating the second term by parts, we conclude that
This must be true for every function \(\varDelta x(t)\), in particular for a function that is equal to 0 everywhere except for a small vicinity of a moment \(t\), For this function, the integral is proportional to the value of the expression \(\displaystyle \frac{\partial L}{\partial x} -\displaystyle \frac{d}{dt}\left( \displaystyle \frac{\partial L}{\partial \dot{x}}\right) \) at the point \(t\). Since the integral is 0, this expression must also be equal to 0:
The resulting equations are known as Euler-Lagrange equations.
Case of Newton’s laws. In particular, for the Newton’s case, when
for each of the components \(x_i(t)\), we have \(\displaystyle \frac{\partial L}{\partial x_i}=\displaystyle \frac{\partial V}{\partial x_i}\) and \(\displaystyle \frac{\partial L}{\partial \dot{x}_i}=-m\cdot \displaystyle \frac{dx_i}{dt}.\) Thus, Euler-Lagrange’s equations lead to \(\displaystyle \frac{\partial V}{\partial x}+m\cdot \displaystyle \frac{d}{dt}\left( \displaystyle \frac{dx_i}{dt}\right) =0,\) i.e., to Newton’s equations \(m\cdot \displaystyle \frac{d^2x_i}{dt^2}=-\displaystyle \frac{\partial V}{\partial x_i}.\)
General case. In the general case, Euler-Lagrange equations take the form \(\displaystyle \frac{\partial L}{\partial \varphi }-\sum \limits _{i=1}^3 \displaystyle \frac{\partial }{\partial x_i}\left( \displaystyle \frac{\partial L}{\partial \varphi _{,i}}\right) =0,\) where \(\varphi _{,i}\mathop {=}\limits ^\mathrm{def}\displaystyle \frac{\partial \varphi }{\partial x_i}.\)
Appendix 2: Proof of Proposition 1
\(1^\circ \). Let us first apply the condition \(\mu (x+x_0, a+x_0,\sigma )=\mu (x,a,\sigma )\) with \(x_0=-a\). Then, we get \(\mu (x,a,\sigma )=\mu (x-a,0,\sigma )\), or, equivalently,
where we denoted \(\mu _1(z,\sigma )\mathop {=}\limits ^\mathrm{def}\mu (z,0,\sigma )\).
\(2^\circ \). In terms of the function \(\mu _1\), the condition \(\mu (\lambda \cdot x,\lambda \cdot a,\lambda \cdot \sigma )=\mu (x,a,\sigma )\) takes the form \(\mu _1(\lambda \cdot (x-a),\lambda \cdot \sigma )=\mu _1(x-a,\sigma )\). Let us apply this condition for \(\lambda =\sigma ^{-1}\). Then, we conclude that \(\mu _1(z,\sigma )=\mu _1\left( \displaystyle \frac{z}{\sigma },1\right) \), or, equivalently, \(\mu _1(z,\sigma )=\mu _0\left( \displaystyle \frac{z}{\sigma }\right) \), where we denoted \(\mu _0(z)\mathop {=}\limits ^\mathrm{def}\mu _1(z,1)\).
Substituting this expression for \(\mu _1(z,\sigma )\) in terms of \(\mu _0\) in the expression for \(\mu \) in terms of \(\mu _1\), we conclude that \(\mu (x,a,\sigma )=\mu _0\left( \displaystyle \frac{x-a}{\sigma }\right) .\)
\(3^\circ \). Substituting the expression for \(\mu \) in terms of \(\mu _0\) into the condition \(\mu (-x,-a,\sigma )=\mu (x,a,\sigma )\), we conclude that \(\mu _0(-z)=\mu _0(z)\). Thus, \(\mu (x,a,\sigma )=\mu _0\left( \displaystyle \frac{|x-a|}{\sigma }\right) .\)
\(4^\circ \). For \(a_1=a_2=0\) and \(\sigma _1=\sigma _2=1\), the fusion condition implies that
for some \(a\), \(C\), and \(\sigma \). The left-hand side attains its maximum (\(=\)1) at \(x=0\), the right-hand side attains its maximum (which is equal to \(C\)) for \(x=a\). Since these two sides are one and the same function, we conclude that \(a=0\) and \(C=1\), i.e., that \(\mu _0^2(x)=\mu (k_2\cdot x)\) for some constant \(k_2\) (\(=1/\sigma \)). For an auxiliary function \(\ell (x)\mathop {=}\limits ^\mathrm{def}\ln (\mu _0(x))\) we conclude that \(2\cdot \ell (x)=\ell (k_2\cdot x)\).
Similarly, if we consider 3, 4, etc. terms, we conclude that \(3\cdot \ell (x)=\ell (k_3\cdot x)\), \(4\cdot \ell (x)=\ell (k_4\cdot x)\), etc.
\(4^\circ \). The function \(\mu _0(x)\) for \(x>0\) is monotonously decreasing from 1 to 0. Therefore, \(\ell (x)\) is monotonously decreasing from 0 to \(-\infty \). Since \(\mu \) (and thus, \(\mu _0\)) is continuous, the function \(\ell (x)\) is also continuous, and hence, there exists an inverse function \(i(x)=\ell ^{-1}(x)\), i.e., such a function that \(i(\ell (x))=x\) for every \(x\).
For this inverse function, the equality \(n\cdot \ell (x)=\ell (k_n\cdot x)\) turns into \(i(n\cdot \ell (x))=i(\ell (k_n\cdot x))=k_n\cdot x=k_n\cdot i(\ell (x))\). So, if we denote \(\ell (x)\) by \(X\), we conclude that for every \(n\), there exists a \(k_n\) such that \(i(n\cdot X)=k_n\cdot i(X)\).
If we substitute \(Y=n\cdot X\), we conclude that \(i(Y)=k_n\cdot i\left( \displaystyle \frac{Y}{n}\right) \), and therefore, \(i\left( \displaystyle \frac{Y}{n}\right) =\displaystyle \frac{1}{k_n}\cdot i(Y)\).
From these two equalities, we conclude that \(i\left( \displaystyle \frac{m}{n}\,\cdot \, X\right) =\displaystyle \frac{1}{k_n}\,\cdot \, i(m\cdot X)=\displaystyle \frac{k_m}{k_n}\,\cdot \, i(X)\). So, for every rational number \(r\), there exists a real number \(k(r)\) such that \(i(r\cdot X)=k(r)\cdot i(X)\). Therefore, the ratio \(\displaystyle \frac{i(r\cdot X)}{i(X)}\) is constant for all rational \(r\).
\(5^\circ \). Since \(i(X)\) is a continuous function, and any real number can be represented as a limit of a sequence of rational numbers, we conclude that the ratio \(\displaystyle \frac{i(r\cdot X)}{i(X)}\) is constant for real values of \(r\) as well. Therefore, for every real number \(r\), there exists a \(k(r)\) such that \(i(r\cdot X)=k(r)\cdot i(X)\).
We have thus arrived at a functional equation for which all monotonis solutions are known: they are \(i(X)=A\cdot X^p\) for some \(A\) and \(p\); see, e.g., [1]. Therefore, the inverse function \(\ell (x)\) (\(x>0\)) also takes the similar form \(\ell (x)=B\cdot x^m\) for some \(B\) and \(m\). Taking into consideration that \(\mu _0(x)\) and hence \(\ell (x)\) are even functions, we conclude that \(\ell (x)=B\cdot |x|^m\) for all \(x\).
\(6^\circ \). Now, for every \(a_1>0\), if we take \(a_2=-a_1\) and \(\sigma _1=\sigma _2=1\), then the fusion property implies that \(\mu _0(x-a_1)\cdot \mu _0(x+a_1)=C\cdot \mu _0\left( \displaystyle \frac{x-a}{\sigma }\right) \) for some \(a\) and \(\sigma \). The left-hand side of this equation is an even function, so the right-hand side must also be even, and therefore \(a=0\). So, \(\mu _0(x\,-\,a_1)\mu _0(x\,+\,a_1)=C\cdot \mu _0\left( \displaystyle \frac{x}{\sigma }\right) \). For \(x=0\) we get \(\mu _0(a_1)\cdot \mu _0(a_1)=C\). Turning to logarithms, we conclude that for every \(a_1\), there exists a \(k(a_1)\) (\(=1/\sigma \)) such that \(\ell (x-a_1)+\ell (x\,+\,a_1)=\ell (k(a_1)\cdot x)\,+\,2\cdot \ell (a_1)\). If we substitute here \(\ell (x)=B\cdot |x|^m\), and divide both sides by \(B\), we conclude that \(|x-a_1|^m+|x+a_1|^m=(k(a_1))^m\cdot |x|^m+2\cdot a_1^m\).
Let us show that this equality is satisfied only when \(m=2\).
\(7^\circ \). When \(x>0\), and \(a_1\) is sufficiently small, then \(x+a_1\), \(x\), and \(x-a_1\) are all positive, and, therefore, \((x-a_1)^m+(x+a_1)^m=(k(a_1))^m\cdot x^m+2\cdot a_1^m\). If we move \(2\cdot a_1^m\) to the left-hand side, and divide both sides by \(x^m\), we conclude that \(\left( 1-\displaystyle \frac{a_1}{x}\right) ^m+\left( 1+\displaystyle \frac{a_1}{x}\right) ^m-2\cdot \left( \displaystyle \frac{a_1}{x}\right) ^m=(k(a_1))^m\). The left-hand side of the resulting equality depends only on the ratio \(z=\displaystyle \frac{a_1}{x}\), the right-hand side only on \(a_1\). Therefore, if we choose any positive real number \(\lambda \), and take \(a_1^{\prime }=\lambda \cdot a_1\) and \(x^{\prime }=\lambda \cdot x\) instead of \(a_1\) and \(x\), then we can conclude that the left-hand side will be still the same, and therefore, the right-hand side must be the same, i.e., \((k(a_1))^m=(k(\lambda \cdot a_1))^m\). Since \(\lambda \) was an arbitrary number, we conclude that \(k(a_1)\) does not depend on \(a\) at all, i.e., that \((k(a_1))^m\) is a constant. Let us denote this constant by \(k\).
So the equation takes the form \((1-z)^m+(1+z)^m=k+2\cdot z^m\). When \(z\rightarrow 0\), then the left-hand side tends to 2 and right-hand side to \(k\), so from their equality we conclude that \(k=2\), i.e., that \((1-z)^m+(1+z)^m=2+2\cdot z^m\).
The left-hand side is an analytical function of \(z\) for \(z\) close to 0. Therefore the right-hand side must also be a regular analytical function in the neighborhood of 0 (i.e., it must have a Taylor expansion for \(z=0\)). Hence, \(m\) must be an integer.
The values \(m<2\) are impossible, because for \(m=0\) our equality turns into a false equality \(2=3\), and for \(m=1\) it turns into an equality \(1-z+1+z=2+z\), which is true only for \(z=0\). So \(m\ge 2\).
Since both sides are analytical in \(z\), the second derivatives of both sides at \(z=0\) must be equal to each other. The second derivative of the left-hand side at \(z=0\) is equal to \(m\cdot (m-1)\). The second derivative of the right-hand side is equal to \(2m\cdot (m-1)\cdot z^{m-2}\).
If \(m>2\), then this derivative equals 0 at \(z=0\) and therefore cannot be equal to \(m\cdot (m-1)\). So \(m\ge 2\), and \(m\) cannot be greater than 2. So, \(m=2\). Thus, \(\ell (x)=B\cdot x^2\), and hence \(\mu _0(x)=\exp (-\beta \cdot x^2)\) for some \(\beta >0\). The proposition is proven.
Appendix 3: Proof of Proposition 2
Let us assume that \(s_k\rightarrow x\), and let us prove that in this case, the ratios
also tend to \(x\), i.e., that for every \(\varepsilon >0\), there exists an \(n\) for which, for all \(N\ge n\), we have \(|X_N-x|\le \varepsilon \).
Since \(s_k\rightarrow x\), there exists an integer \(n_0\) such that for all \(k\ge n_0\), we have \(|s_k-x|\le x+\displaystyle \frac{\varepsilon }{2}\). In particular, this means that for such \(k\), we have \(s_k\le x+\displaystyle \frac{\varepsilon }{2}\). We can represent the numerator \(\mathcal{N}\) of the ratio \(X_N\) as
where \(\mathcal{N}_0\mathop {=}\limits ^\mathrm{def}\sum \limits _{k=0}^{n_0} s_k\cdot (s_{k+1}-s_k)^{-2}\). Since \(s_k\le x+\displaystyle \frac{\varepsilon }{2}\), we conclude that
where we denoted \(\varDelta \mathop {=}\limits ^\mathrm{defe }\sum \limits _{k=n_0+1}^{N} (s_{k+1}-s_k)^{-2}\). Similarly, for the denominator \(\mathcal D\) of the ratio \(X_N\), we get an expression \(\mathcal{D}=\mathcal{D}_0+\varDelta \), where
Thus,
The right-hand side of this inequality can be represented as
Here, \(|s_k-x|\le \displaystyle \frac{\varepsilon }{2}\) and \(|s_{k+1}-x|\le \displaystyle \frac{\varepsilon }{2}\) implies that
Thus, \((s_{k+1}-s_k)^{-2}\ge \varepsilon ^{-2}\) and so, \(\varDelta \ge (N-n_0)\cdot \varepsilon ^{-2}\). When \(N\rightarrow \infty \), we have \(\varDelta \rightarrow \infty \) and thus,
for sufficiently large \(N\). For such \(N\), we get \(X_n=\displaystyle \frac{\mathcal{N}_0}{\mathcal{D}_0}\le x+\displaystyle \frac{\varepsilon }{2}+\displaystyle \frac{\varepsilon }{2}=x+\varepsilon \). Similarly, for sufficiently large \(N\), we get \(X_N\ge x-\varepsilon \). The proposition is proven.
Appendix 4: Example: Applying Formula (6) to the Divergent Geometric Series \(\sum z^i\) for \(|z|\ge 1\)
When \(|z|>1\), the series \(\sum z^i\) diverges. Here, \(s_1=1\), \(s_2=1+z\), ..., and, in general, \(s_k=1+z+\cdots +z^k=\displaystyle \frac{z^{k+1}-1}{z-1}.\) Thus, \(s_{k+1}-s_k=\displaystyle \frac{z^{k+2}-z^{k+1}}{z-1}=z^{k+1}\). So, the denominator \(\mathcal D\) of the formula (6) has the form \(\mathcal{D}=\sum \limits _{k=0}^N z^{-2\cdot (k+1)}\). In the limit, when \(N\rightarrow \infty \), we get \(\mathcal{D}\rightarrow \displaystyle \frac{z^{-2}}{1-z^{-2}}\).
For the numerator, we similarly have
In the limit, when \(N\rightarrow \infty \), we get \(\mathcal{N}\rightarrow \displaystyle \frac{1}{z-1}\cdot \left( \displaystyle \frac{z^{-1}}{1-z^{-1}} -\displaystyle \frac{z^{-2}}{1-z^{-2}}\right) .\) Thus,
Here,
Therefore,
Adding two fractions in parentheses, we get
Thus,
The terms \(z^{-1}\), \(z^{-1}\), and \(z^{-2}\) cancel each other, as well as the terms \(1-z^{-2}\) in the numerator and in the denominator. Thus, we get \(x=\displaystyle \frac{1}{1-z^{-1}}\).
For example, for \(z=2\), we get \(x=1+2+4+\cdots =\displaystyle \frac{1}{1-1/2}=2\).
Rights and permissions
Copyright information
© 2015 Springer International Publishing Switzerland
About this chapter
Cite this chapter
Kosheleva, O., Reiser, R., Kreinovich, V. (2015). Formalizing the Informal, Precisiating the Imprecise: How Fuzzy Logic Can Help Mathematicians and Physicists by Formalizing Their Intuitive Ideas. In: Seising, R., Trillas, E., Kacprzyk, J. (eds) Towards the Future of Fuzzy Logic. Studies in Fuzziness and Soft Computing, vol 325. Springer, Cham. https://doi.org/10.1007/978-3-319-18750-1_14
Download citation
DOI: https://doi.org/10.1007/978-3-319-18750-1_14
Published:
Publisher Name: Springer, Cham
Print ISBN: 978-3-319-18749-5
Online ISBN: 978-3-319-18750-1
eBook Packages: EngineeringEngineering (R0)