Radial Basis Function and Its Application for Engineering Field

Radial Basis Function and Its Application for Engineering Field

Kittali Roshitha

Vignans Institute of Engineering for Women (A), Department of IT, 530049 Visakhapatnam, India

Tekumalla Sai Sanjana

Vignans Institute of Engineering for Women (A), Department of IT, 530049 Visakhapatnam, India

Abstract – An essential part of studying mathematical models in science and engineering is the study of partial differential equations. To solve differential equations, academics have created a variety of tools and methods. It has been demonstrated that radial basis functions are an effective basis function for estimating the solutions of partial and ordinary differential equations. Researchers have created a variety of radial basis function techniques to solve a number of well-known differential equations. It has been created to approximate the solution using a variety of techniques that result in the creation of hybrid methods. Because radial basis function approaches can handle meshless domains, they are frequently employed in numerical analysis and statistics. The various radial basis function approaches were examined in this work, with an emphasis on the methods being used to determine the value of the shape parameter. The possible shape parameters to obtain the optimal value of the numerical solutions are examined, as are the mathematical formulations of the various radial basis function approaches. The current effort will establish a platform for comprehending the evolution of radial basis functions, which may play a significant role in the subsequent development of the approach.

Keywords – Partition of unity, Shape parameter, Partial differential equation, Kansa collocation method, Radial basis function

1. INTRODUCTION

   Radial basis functions (RBFs), which approach the numerical solution when implemented and make the process computationally efficient, are an efficient numerical method for solving partial differential equations (PDEs). Initially designed to solve interpolation matrices, the radial basis function was eventually used to solve partial differential equations. The fact that RBF is simple to use and performs well in dynamic and irregular domains is one of its key features. RBF techniques are widely used in statistics and numerical analysis. These applications include, but are not limited to, numerical solutions of PDEs, geomodelling, machine learning, price options, neural networks, data mining, and image processing.

   There are several numerical techniques that are available to solve a modeled partial differential equation. But each method has its own advantage and disadvantage, such as The initial difference method is a straightforward and practical approach. However, it becomes quite complicated for irregular domains;

   the finite volume method is widely used in computational fluid dynamics, surface integral over control volumes, but has complexity involved; the finite element method is most popular because it is most flexible over complex domains but requires a lot of integration. The quadrature technique [1], B- spline finite element methods [2], RBF methods [3], exponential B-spline with PSO [4], and the modified cubic B- spline differential quadrature method [5, 6] are some of the computational techniques that some researchers have developed based on the above defined techniques that have been implemented for finding the numerical solutions of PDEs. Other methods include the RBF collocation method as pseudo-spectral methods, Kansa’s method for solving parabolic, elliptic, and hyperbolic PDEs, and the finite difference method [710]. The most challenging area of research nowadays is the numerical solutions of fractional ordinary equations (ODEs) and PDEs, for which numerous researchers have developed a variety of numerical techniques. Arora G. [12] introduced the residual power series method for fractional relaxationoscillation equations, while Maayah [11] proposed the multistep Laplace optimized decomposition method for fractional systems of ODEs. The efficiency of the proposed method was tested using the RungeKutta method of order four. The Dirichlet model was introduced by Arqub [13] using a computational method based on the time-fractional reproducing kernel. An overview of recent studies [14] on the theoretical and experimental investigation of the thermal conductivity of several nanofluids is given in this article. The current study looks at several factors, such as temperature, solid volume percentage, type, size, shape, magnetic field, pH, ultrasonic time, and surfactant, that significantly affect a nanofluid’s capacity to transport heat. Additionally, various attractive and tenable hypotheses that improve nanofluids’ thermal conductivity are discussed. Brownian motion, thermophoresis, nanoclustering, interfacial nano-layer, and osmophoresis are the last key heat transmission mechanisms that significantly contribute to increasing thermal conductivity. As a result, the characteristics of nanofluids’ heat transmission are taken into account. Additionally, various numerical methods for solving fluid flow problems were explored. The structure of the system integrates a number of MCES components, which includes as resource hubs, distributed power production, as well as storage systems, as well as is put into practice as a series of case studies to show how well it works [15] This technique aims to improve system

   dependability by tackling the inherent uncertainty associated with green energy supplies. The chief aim of this effort is to speech the complex issue of power organization fashionable a smart city, including the doubts related with renewable energy sources then the incorporation of demand response programs (DRPs). To tackle these difficulties, the paper utilizes the integrated inside the Python program, while doubts fashionable the scheme is demonstrated using the Monte Carlo technique. The findings indicate that accounting for the variability of green energy supplies may decrease the overall system cost by 504 $ via the use of Demand Response Programs (DRP) [16]. The Kansa method, the Hermite symmetric approach, the localized method, and the hybrid method are some of the generalized RBF approaches we study in this paper. We also talked about how mesh-free solutions like RBF are preferred over mesh-based ones. Also mentioned is a recent advancement in RBF approximation for PDE solution. [17]. This paper presents an efficient full approximation scheme-full multilevel (FAS-FML) algorithm for the radial basis function-based finite difference (RBF-FD) method. The algorithm produces an accurate solution by solving the discretized equations from the coarsest level to the coarser level and then to the finer level to desire the finest level node points [18]. In this article, the problem of a two- phase unsteady boundary layer flow past an inclined stretched surface has been considered. The impact of electrification and transverse force has been analyzed in presence of Buoyancy force [19] The research intends to contribute to the understanding of the interplay between these factors, elucidating their combined effects on heat transfer efficiency, boundary layer behavior, and velocity distribution within the nanofluid flow. By analyzing these interactions, researchers seek to identify optimal conditions for maximizing heat transfer rates, crucial for various engineering and industrial applications where efficient heat dissipation or transfer is paramount. This comprehensive analysis bridges theoretical considerations with practical implications, offering valuable insights for designing and optimiing systems involving nanofluid under magnetic field influence and porous surfaces.

   [20] Gandi et al. proposed an IoT-based sewer vault monitoring and cautioning system that continuously monitors environmental parameters and issues alerts during hazardous situations [21]. AI techniques have also demonstrated strong potential in healthcare diagnostics, particularly in medical image analysis. Deep learning models, especially convolutional neural networks (CNNs), have been applied to detect respiratory diseases from chest X-ray images, enabling faster, more accurate diagnosis of pneumonia and other pandemic pathogens [22]. Similarly, AI-driven systems are being used in educational platforms to improve student employability through intelligent learning management systems that provide skill assessment, career guidance, and predictive placement analysis [23]. In industrial applications, AI-based predictive maintenance frameworks integrate techniques such as YOLOv5, Random Forest regression, and neural networks to monitor aircraft components and predict remaining useful life of critical systems [24].

   AI-based sensor systems are also used in safety applications, such as smart borewell rescue systems that detect accidents,

   automatically alert emergency services, and activate rescue mechanisms [25].

2. RADIAL BASIS FUNCTION (RBF)
   1. Radial basis function

      Radial basis function A function : t is called radial if there exist a function of one variable: [0,) R such that

      (x) = (x), here Euclidean norm . is used and t N. (r) is a univariate continuous real valued radial basis function whose value based upon distance value that is measure from any fixed centre point or the origin [14].

      It is evident from the definition that is a unique function that solely depends on the distance between points and is radially symmetric. Since the interpolation problem is agnostic to the space dimension, applying the radial basis function to the high dimensional problem is simple. In any dimension of space, one can deal with

      the univariate function rather than a multivariate function . Our focus is on radial basis function types that are characterized by piecewise smooth RBFs that are indefinitely differentiable, free of the shape parameter , and have parameters known as the shape parameter.

   2. RBF types

      RBFs come in a variety of forms. The following are a few acknowledged RBFs:

      1. infinitely smooth RBFs: These RBFs are based on the shape parameter > 0, which regulates the RBF’s outline or shape. The shape of RBFs becomes flat if tends to 0.The various kinds of infinitely smooth RBFs are shown in Table 1.
      2. RBFs that are piecewise smooth; these RBFs lack a shape parameter. Table 2 displays many kinds of piecewise smooth RBFs.
3. PREPARE YOUR PAPER BEFORE STYLING

   The various types of radial basis functions (RBFs) that were previously covered are summarized in Fig. 1. Examining Figs. 2 and 3 makes it evident that variations in the form parameter significantly affect the profiles of the radial functions. Fig. 2 displays the Gaussian RBF for form parameter values of 1, 0.5, and 0.2, showing how the function flattens as the parameter decreases. Similarly, Fig. 3 highlights the corresponding changes in the function’s curvature by displaying the inverse multiquadric RBF for shape parameter values of 1, 0.5, and 0. RBF techniques for PDE solutions RBF techniques are renowned for their simplicity and ease of use when approximating multivariate scattered data. The following is a current historical and chronological development plan (Fig. 4) of RBF methods for solving partial differential equations:

   PDE solutions using the Kansa collocation method The Kansa method, also referred to as the RBF collocation method, is one mesh-free technique. Mesh-free approaches provide many advantages over mesh methods. Because they don’t need domain or surface discretisation, they are less expensive. An asymmetric technique was introduced by Kansa. An RBF-based method for resolving PDEs is the Kansa methodology .Mathematically, take x and in Rd, d N, consider the norm that is, nodes that are spotted arbitrarily in the domain ² d and give each x a collection of neighbourhood nodes xi that are integrated in the supportive domain in the following way:

   The following is a summary of the Kansa collocation method:

   1. 1. 1. Examine a PDE on a certain domain with boundary conditions.
         2. Assume that a linear combination of RBFs with node points is its solution.
         3. Implement the assumed solution at given equation and boundary conditions.
         4. Resultant in the form of linear system of equations.

            Because the Kansa methods are used to solve different PDEs, they have drawbacks. This method’s primary drawback is its extremely high computing cost, which is caused by the unsymmetric interpolation matrix. In the domain nearest to the boundary, this method’s accuracy is lower. Raising the interpolation points that result in a high condition number matrix is the most straightforward method for improving accuracy and reducing errors. However, the resulting matrix becomes ill-conditioned by increasing the points that are now taken throughout the entire domain.

            Fig. 1 Various types of RBFs

            where i is unknown coefficients and N represents the numbers of node points. By substituting this solution u(x) in PDE gives the linear system of equations as

            AX = B where X = [(x1), (x2), (x3), (xN )]

            Please take note of the following items when proofreading spelling and grammar: Symmetric collocation method (SCM)

            1. Symmetric collocation method (SCM)

               After modification in the Kansa method, a new method comes in existence, which is known as Symmetric collocation method. This method is based upon Hermite interpolation and proposed by Fasshauer and also invent the RBF expansion for approximating the function. After applying the collocation conditions, there is a requirement of a non-singular symmetric collocation matrix. Symmetric and non-symmetric techniques had been applied for different applications. These methods are compared by Power & Barraco and find the result as the symmetric collocation technique is surpassing the non- symmetric (Kansa method) technique due to the lower computational cost. But the implementation of Kansa scheme is unproblematic. The symmetric collocation method is also used by Leitao to solve 2-dimensional electrostatic problems.

            2. Modified collocation method (MCM)

               The interpolation matrix is symmetric because the modified collocation method is an improved version of the symmetric approach. Chen proposed the improved collocation approach, which makes advantage of Green second identity.

               The primary issue in applying the Kansa technique to determine the solutions of the several PDEs is the ill- conditioned interpolate matrix. Many approaches were proposed to address the problem, including the domain decomposition method, compactly supported RBF, and pre- conditioning. Pre-conditioning is the process of converting a collection of linear equations into a new system that is a useful method for iterative solution. The term “pre-conditioner” refers to the transformation that a matrix produces.

            3. RBF methods with shape parameters

   The Optimizing the shape parameter is a constant focus in the realm of RBF research. Numerous investigations have been carried out in this area. Table 3 lists several techniques for determining the optimal form parameter. To ascertain the relationship between the value of and the accuracy of thesolutions in their study, Huang et al. employed arbitrary precision computing. In order to prevent the singularity caused by round-off error that arises in regular 16- digit precision arithmetic when the parameter value is tiny, their research suggests using 100-digit precision arithmetic to discover the solution of radial basis functions. Based on the gathered numerical data, they developed incorrect formulations for the value of and grid spacing.

   Similar to the results of the finite element method, other approaches have or have not provided the accuracy of the optimal method that was reached. Verifying the procedure’s accuracy and strength in comparison to others is the primary goal. In the method described, an iteration algorithm is provided by computing a matrix that is obtained by the multi- quadric functions outer side of the time loop in which the boundary operator B on the MQ basis function on and matrix W over domain \ define the value of the operator L on the MQ basis function. The MQ kernel employed an easy-to-use shape parameter, and the value of this parameter determines how accurate the outcome is. The second derivative, which has a shape parameter that influences the correctness of the PDE solution, is approximated using the RBF finite difference (RBF-FD) method, which is based on MQ and is calculated in a polynomial structure. By eliminating the leading error term of

   the RBF-FD method, the optimal value of the form parameter is determined, increasing convergence speed and improving solution correctness. The optimal shape parameter is found by combining finite difference and compact differencing approaches.

   The RBF approximation can generate a suitable estimate for a large collection of data points that yields a smooth result for specific tangled points when the RBFs are nearly flat and the shape parameter value is chosen appropriately. The inverse multi-quadric (IMQ) RBF function was used in research by Kazeem et al. to write and develop a method for solving partial differential equations. The choice of form parameters is prioritized and needs careful consideration. The method is an algorithm that selects the ideal form parameter with the minimum root mean square error (RMSE) after conducting a number of interpolation tests while varying the range of the shape parameters. Cavoretto and Rossi devised an algorithm for spherical interpolation to discover the numerical solution of a problem using a basis function that further projected a technique utilizing the partition of unity approach. The spherical radial basis function is primarily used in local approximation by the author. Numerous procedures can be carried out in this method in an equivalent manner. In an extension of this work, Cavoretto and Rossi proposed a partition of unity algorithm that divides the domain into nodes or cells. Cell search is the main method used in this process. Additionally, the author used a cube partition finding approach to extend this two-dimensional algorithm to three dimensions. Safdari et al looked into the division of unity method’s applications for solving parabolic partial differential equations.

   The region is divided into intersecting local domains using the division of unity approach. This method is crucial for selecting a family of continuous functions that are compactly supported. The RBF-PUM approach is the most effective strategy to get improved accuracy while lowering computing costs. This method’s primary benefits in high-dimensional situations are its ability to maintain geometrical flexibility, reduce computation costs, and enable adaptive approximation. Using weight functions, which determine the manner of partitioning unity, the local approximation is defined on sub- domains and then merged to form the global approximation.

4. CONCLUSION

An overview of the radial basis function and methods based on RBFs for solving different PDEs is given in this article. We try to Describe some recent advancements in RBF techniques and strategies for determining the shape parameter’s optimal value. The purpose of this paper is to acquaint the reader with RBF approaches as collocation methods, global approximation, RBF-DQ method, RBF-PUM, and local collocation RBF approaches. These techniques are very helpful when solving large-scale problems and help reduce processing expenses. For smooth RBFs, the form parameter’s smallest value results in acceptable accuracy, but the radial basis’s near flatness causes the interpolation matrices to be poorly conditioned. For smooth RBFs, the form parameter’s lowest value results in acceptable accuracy, but the radial basis’s near flatness causes the interpolation matrices to be poorly conditioned. This paper lists a number of algorithms that were proposed to address this problem. By examining the ideal shape parameter value for increased accuracy and stability of RBF approximations, the existing methods can be further enhanced. Research is still

being done to determine how effective RBF methods are at solving higher-order PDEs.

REFERENCES

1. 1. Arora, G., Joshi, V.: A computational approach for one and two dimensional Fishers equation using quadrature technique. Am. J. Math. Manag. Sci. 40(2), 145162 (2021).

   https://doi.org/10.1080/01966324.2021.1933660

2. 2. Krl, E., Irk, D.: Efficient techniques for numerical solutions of Fishers equation using B-spline finite element methods. Comp. Appl.

   Math. 42, 151 (2023). https://doi.org/10.1007/s40314-023-02292

3. 3. He, Y., Sun, X., Huang, P., Xu, H.: Evaluation of automatic algorithm for solving differential equations of plane problems based on BP neural network algorithm. J. Intell. Fuzzy Syst. 38(2), 1593602 (2020). https://doi.org/10.1002/ num.22980
4. Rani, R., Arora, G., Emadifar, H., Khademi, M.: Numerical simulation of one-dimensional nonlinear Schrodinger equation using PSO with exponential B-spline. Alex. Eng. J. 79, 644651 (2023)
5. Arora, G., Mishra, S., Emaifar, H., & Khademi, M. Numerical simulation and dynamics of Burgers equation using the modified cubic B-spline differential quadrature method. Discret. Dyn. Nat. Soc. 2023, 18 (2023). https://doi.org/10. 1155/2023/5102374
6. Arora, G., Joshi, V., Mittal, R.C.: A spline-based differential quadrature approach to solve sine-gordon equation in one and two dimension.

   Fractals 30(07), 2250153 (2022) 7

7. .Ahmad, A.E.: Local formulation for time-dependent PDEs. MDPI 216 (2019). https://doi.org/10.3390/math7030216 8.
8. Kansa, E.J.: Multiquadricsa scattered data approximation scheme with applications to computational fluiddynamicsII solutions to parabolic, hyperbolic and elliptic partial differential equations. Comput.

   Math. Appl. 19

9. 147161 (1990) 9. Fasshauer, G.E.: RBF collocation methods as pseudospectral methods, WIT transactions on modelling and simulation,

   Southampton, UK, (2005), p 39

10. Arora, G., Bhatia, G.S.: A meshfree numerical technique based on

    radial basis function pseudospectral method for Fishers equation.

    IJNSNS (2019). https://doi.org/10.1515/ijnsns-2018-0091

11. Maayah, B., Moussaoui, A., Bushnaq, S., Abu Arqub, O.: The multistep Laplace optimized decomposition method for solving fractional-order coronavirus disease model (COVID-19) via the Caputo fractional approach. Demonstratio Math. 55(1), 963977 (2022)
12. Arora, G., Pant, R., Emadifar, H., Khademi, M.: Numerical solution of fractional relaxationoscillation equation by using residual power series method. Alex. Eng. J. 73, 249257 (2023)
13. Arqub, O.A., Maayah, B.: Adaptive the Dirichet model of mobile/immobile advection/dispersion in a time-fractional sense with the reproducing kernel computational approach: Formulations and approximations. Int. J. Mod. Phys. B 37(18), 2350179 (2023)
14. Mishra, Debasish, and T. N. Samantara. “Current advancements in fluid- flow problems: A brief review.” World Journal of Advanced Research and Reviews 21.1 (2024): 393-398.
15. Sahoo S. S., Garg A., Ansari M. F., Niazi K. R. and Mishra D., “Operational Analysis of Multi-Carrier Energy Systems Under

    Uncertainty and Integration of Renewable Energy and Storage,” 2025 International Conference on Power Electronics and Energy (ICPEE), Bhubaneswar, India, 2025, pp.1-6, doi: 10.1109/ICPEE65973.2025.11411352.

16. Garg A., Mishra D, Sahoo S. S. and Niazi K. R., “Operational Analysis of Multi-Carrier Energy Systems Under Uncertainty and Demand Response Integration,” 2025 International Conference on Electrical, Electronics, and Computer Science with Advance Power Technologies – A Future Trends (ICE2CPT), Jamshedpur, India, 2025, pp. 1-5, doi: 10.1109/ICE2CPT66440.2025.11340180.
17. Mishra, D., Barik, N.B., Samantara, T. (2026). A Radial Basis Function and Its Application in Solving Partial Differential EquationsA Brief Review. In: Das, K.N., Biswas, P., Budinska, I., Mallipeddi, R. (eds) Proceedings of the International Conference on Computational Mathematics and Applications. ICCMA 2025. Lecture Notes in Networks and Systems, vol 1520. Springer, Singapore. https://doi.org/10.1007/978-981-96-8987-3_8
18. Mishra Debasish, Barik Nikunja Bihari, Ojha Kanaka Lata, samantara Tumbanath, Sekhar T. V. S.. (2025). Meshfree Full Approximation Scheme-Full multilevel algorithm for Convection-dominated problems. Journal of Computational Analysis and Applications (JoCAAA), 34(12), 910925. Retrieved from https://eudoxuspress.com/index.php/pub/article/view/4631
19. D. Mishra and T. Samantara, Modeling of an Unsteady Two-Phase Flow Over a Linear Inclined Stretching Sheet with Consideration of Transverse Force and Electrification of Particles, JP Journal of Heat and Mass Transfer, vol. 38, no. 6, pp. 953975, 2025.
20. Mishra, D., & Samantara, T. (2025). Analyzing Combined Effects of Inclination Angle Stretching Sheet with Porous Media On Fe3O4+C2H6O2 Nanofluid Flow. Journal of Advanced Research in Numerical Heat Transfer, 35(1), 6887. https://doi.org/10.37934/arnht.35.1.6887
21. [1] N. Gandi, J. S. Karri, S. S. S. Nune, L. Ravada and D. Darapu, Enhanced Urban Infrastructure: IoT-based Sewer Vault Monitoring and Cautioning System, 2024 Second International Conference on Inventive Computing and Informatics (ICICI), Bangalore, India, 2024,

    pp. 552555, doi:10.1109/ICICI62254.2024.00096.

22. [2] N. Gandi, S. S. S. Damisetti, K. G. S. Bharathi, D. M. Santhoshini and V. Pavani, Future Diagnosis: Detection of Pneumonia and Pandemic Pathogen using CNN, 2024 3rd International Conference on Applied Artificial Intelligence and Computing (ICAAIC), Salem, India, 2024, pp. 156160, doi:10.1109/ICAAIC60222.2024.10575804.
23. [3] N. Gandi et al., AI-Driven Placement LMS Portal for Enhanced Student Employability, in International Conference on Computer Science and Communication Engineering, Atlantis Press, 2025. doi: 10.2991/978-94-6463-858-5_3.
24. [4] N. Gandi et al., Integrated AI-Driven Aircraft Maintenance System with Real-Time Crack Detection, Battery Life Estimation, Jet Engine Predictive Maintenance, in Proceedings of the International Conference on Computer Science and Communication Engineering, Atlantis Press, 2025. doi: 10.2991/978-94-6463-858-5_32.
25. [5] P. Bethapudi and N. Gandi, Smart Rescue System from Bore Well, Journal of Advanced Research in Dynamical and Control Systems, vol. 11, Special Issue, pp. 288296, 2019.