_{1}

Previous multifractal spectrum theories can only reflect that an object is multifractal and few explicit expressions of f( α) can be obtained for the practical application of nonlinearity measure. In this paper, an analytical model for multifractal systems is developed by combining and improving the Jake model, Tyler fractal model and Gompertz curve, which allows one to obtain explicit expressions of a multifractal spectrum. The results show that the model can deal with many classical multifractal examples well, such as soil particle-size distributions, non-standard Sierpinski carpet and three-piece-fractal market price oscillations. Applied to the soil physics, the model can effectively predict the cumulative mass of particles across the entire range of soil textural classes.

A multifractal system is a generalization of a fractal system in which a single exponent (the fractal dimension) is not enough to describe its dynamics; instead, a continuous spectrum of exponents is needed [_{q} vs. q [_{q} is the generalized dimension for a dataset and q is an arbitrary set of exponents. Another useful multifractal spectrum is the graph of f(α) vs. α [_{q} can be linked together by the Legendre transformation.

However, the direct determination of a statistically stable and accurate f(α) from real world data is often considered problematic because it demands large amounts of data samples, besides being susceptible to large errors due to logarithmic corrections [_{q} vs. q or f(α) vs. α, which are only the graphs and do not yield explicit expressions for the functions describing the behavior of a multifractal system [

In this paper, an analytical model for multifractal systems is developed by combining and improving the Jaky model, Tyler fractal model and Gompertz curve, which allows one to obtain explicit expressions of a multifractal spectrum.

Jaky [

where F_{2} is the cumulative mass of particles with equivalent diameter ≤ d (mm); p is the index characterizing the stretching of the curve; and d_{0} is the largest diameter (mm).

The Jake model produces a sigmoidal shape similar to the left-hand side of a Gaussian lognormal distribution. This model was first introduced into the soil science literature from the geotechnical discipline by Buchan et al. [

Mandelbrot first established the fractal PSD model for two-dimensional space [

where A(r > R) is the cumulative mass of grain sizes “r” larger than a specific measuring scale R; C_{a} and λ_{a} are constants relating to the shape factors and total range of scale; and D is the fractal dimension. Tyler and Wheat craft established the fractal PSD model for three-dimensional space based on Mandelbrot’s study [

where M_{T} is the total mass of particles and D is the fractal dimension for particles. In order to compare Equation (1) and Equation (3), they are standardized as:

_{1} and F_{2} distributed on the USDA textural triangle. The red solid lines repre- sent the loci of soils whose mass distribution follows Equation (4) and Equation (5) exactly. However, well- graded soils, such as a silty loam, are not well described by Equation (4) and Equation (5) since they do not fall near the red lines. Thus, it is clear that soil texture affects the performance of the one-parameter model. Given these restrictions for some soil textures, it becomes apparent that difficulties may arise in using the one-para- meter model as the input for quantitative simulation of soil structure and hydraulic properties.

The expression forms of Equation (4) and Equation (5) are very similar, differing only in their exponent. This provides a very useful insight that if the exponent is set to a variable parameter, we can enhance the Jake PSD model to cover all the soil textural classes. The enhanced Jake model (named “Jake-Jun” model based on two researchers’ name: the author of the Jake Model, “Jake”, and the author of this paper, “Jun”, for convenience) proposed by this study has two parameters and is expressed as:

where F(r ≤ d) is the cumulative mass of particles with equivalent diameter ≤ d; d_{0} is the largest diameter; m is an exponential factor that is the extension of the Jake model exponential form; and D is the fragmentation fractal dimension incorporated from the Tyler fractal model. The Jake-Jun model can be utilized for the entire textural triangle for predicting the cumulative mass of particles across the entire range of soil textural classes (

For Equation (6), we assume that the complex fractal (m > 1) derived from natural evolution of simple fractals (m = 1) remains unchanged in fractal dimension D during soil developmental evolution, then

To introduce a time variable t, we set

which corresponds to

Equation (7) is the well-known Gompertz curve. A Gompertz curve, named after Benjamin Gompertz, is a sigmoid function, such as a growth curve. It is a type of mathematical model for a time series, where growth is slowest at the start and end of a period.

The lognormal cascade model that the probability density function of data set successive increments at different time scales was introduced to study of fully developed turbulence [

The cumulative distribution of lognormal distribution function is

Jake-Jun model is the cumulative mass distribution function of particles. We can get its probability density function (particle mass of x layer) by first derivative of F(x):

When m = 2, comparing Equation (10) and Equation (12), to set

Equation (12) can be normalized to

From this, Jake-Jun model is the variations and combinations of a lognormal distribution function. This expansion have two aspects: one the hand, the exponent “2” in a lognormal distribution function becomes the variable parameter m; On the other hand, the amplitude is combined with a logarithm function. To understand parameters m and D, we produced their contour maps (

A mass distribution may be spread over a region in such a way that the concentration of mass is highly irregular. There are two basic approaches to multifractal analysis: fine theory, where we examine the structure and dimensions of the fractals that arise themselves, and coarse theory, where we consider the irregularities of distribution of the measure of balls of small but positive radius r and then take a limit as r → 0 [

Compared to a unifractal, a multifractal can be understood as the more general relationships that the dimension of a unifractal is also a function of the scale of observation. In the Jake-Jun model, we set:

With Equation (6), we get

Equation (16) is the explicit expression of multifractal dimension in Jake-Jun model. When m = 1, D_{k} = D is the Hausdorff dimension of a unifractal. To set k = 2^{−}^{n}, _{k} and ln (1/k) with a series values of m.

Jake-Jun model thinks of a multifractal as the variation of a unifractal. Its basis is a unifractal with Hausdorff dimension D and it uses the variation parameter m to simulate a multifractal. Therefore, the explicit expression (Equation (16)) of multifractal spectrum has parameters D, m and observation scale k.

The Jake-Jun model isn’t just a PSD model, but a new multifractal system. In the multifractal spectrum of f(α) vs. α, the respective measure of the ith cell at every size scale ε is defined by m_{i} = ε^{α} and the number of cells N(m_{i}) with singularity strength falling within α given α and α + dα is considered as _{i}. Let

The cumulative mass distribution function of m_{i} is:

When Equation (18) is regarded a continuous function, such that

According to Equation (12) and Equation (19), we have

Therefore, the explicit expressions of f(α) using Jake-Jun model is

where

When Equation (18) is regarded a discrete function, such that

The Sierpinski carpet (

The construction of the Sierpinski carpet begins with a square which length is L. The square is cut into b^{2} congruent sub squares in a b-by-b grid, and K sub squares are removed. The same procedure is then applied recursively to the remaining C sub squares, ad infinitum. In standard Sierpinski carpet, the same K is used for each iteration and it is a single fractal. Set

When m = 1 in Equation (6), it is a unifractal, then

If we use different K for iteration and then it will be a multifractal Sierpinski carpet. We set

Using Equation (6), we get:

Given the parameters D and m, C_{n} and can be calculated:

Therefore, the multifractal mechanics of Jake-Jun model can be visualized demonstration using the Sierpinski carpet (

Mandelbrot [

Three-piece-fractal generator can be interpolated repeatedly into each piece of subsequent charts. The pattern that emerges increasingly resembles market price oscillations (

Mandelbrot’s Three-piece-fractal is that these self-affine fractal curves exhibit a wealth of structure―a foundation of both fractal geometry and the theory of chaos. However, Mandelbrot did not give a quantitative description of these self-affine fractal curves. The surprise is that Jake-Jun model can finish the task. If we think these segments of the Three-piece-fractal multifractal as particles (

In conclusion, an analytical model (named Jake-Jun model) for multifractal systems was developed by combining and improving the Jake model, Tyler fractal model and Gompertz curve. Previous multifractal theories fail to solve the crucial problem, “How can this nonlinearity measure be used?”, because few explicit expressions of f(α) can be obtained. The Jake-Jun model has solved the crucial problem using the mass cumulative distribution function. The Jake-Jun model is able to deal with many classical multifractal examples well, such as soil particle-size distributions, two-scale Cantor set, non-standard Sierpinski carpet and three-piece-fractal market price oscillations. It is an accurate and simple approach for modeling multifractal systems from experimental data. The Jake-Jun model would be able to apply in soil hydraulic, rainfall distribution, basin structure and in many other multifractal systems.

Jun Li, (2016) An Analytical Model for Multifractal Systems. Journal of Applied Mathematics and Physics,04,1192-1201. doi: 10.4236/jamp.2016.47124