Optimal design of soil dynamic compaction using genetic algorithm and fuzzy system
Introduction
In recent years, the scarcity of land space necessary for new development has prompted a significant interest from local authorities in using various undesirable soils. Dynamic compaction method has been found to be useful to improve the mechanical behavior of underlying of soil layers, especially loose granular materials, in many projects [1], [2], [3], [4], [5], [6], [7], [8], [9]. This method involves the repeated impact of high-energy impacts to the soil surface using steel or concrete tampers with weight ranging 10–40 ton and with drop heights ranging 10–40 m.
An appropriate estimation of dynamic compaction level is a major task for geotechnical engineers, in particular, if multilayer ground is encountered. Traditional methods are unable to estimate this level accurately and numerical methods may be powerful tools to deal with the complexity of this phenomenon. This research develops a fuzzy base method for dynamic compaction analysis. This method, as human brain, can interpret imprecise and incomplete sensory information and provides a systematic method to deal with the corresponding information linguistically. In this model, the input variables are tamper weight, height of tamper drop, print spacing, tamper radius, number of impact and soil layer geotechnical properties. The output involve depth of improvement which verified by database compiled from field tests and other researches. The proposed method is potentially useful for evaluation of improvement of loose granular soils due to dynamic compaction.
Genetic algorithm is a powerful and broadly applicable stochastic search and optimization techniques based on principles from the evolution theory. It has been shown to be suitably robust for a wide variety of problems [10], [11], [12]. This paper also uses this technique to show its abilities in optimizing the dynamic compaction design. This paper also compares different manner of this algorithm and then suggests the optimized structure of genetic algorithm for dynamic compaction phenomenon. Finally a procedure is proposed to achieve the greatest depth of soil improvement upon using dynamic compaction method. The recommended procedure based on the fuzzy modeling considers all involving parameters in dynamic compaction, some of which are tamper weight, dropped height, number of impact, tamper radius, print spacing and soil strength parameters. The best combination of these parameters will be introduced.
Section snippets
Fuzzy set concepts and modeling methods
An inexact phenomenon such as the mental process would effectively be analyzed by using a system such as fuzzy system or neural systems. In this paper, the learning structure analysis system and its practical case study and the effectiveness are explained using fuzzy system. This system is based on fuzzy logic which simulates human's comprehensive inference and judgement and solves the problems of fuzzy information processing which is hard to solve by normal methods. Fuzzy system has been
Genetic algorithm search
The genetic algorithm is a method for solving optimization problems that is based on natural selection, the process that drives biological evolution. In the past few years, the genetic algorithm community has turned much of its attention toward the optimization problems of industrial engineering.
The main differences between GAs and conventional optimization procedure which make GAs more amenable for application to a wide variety of problems are described below.
GAs work by directly manipulating
GA implementation
Fitness function of GA program in this paper is the fuzzy system of dynamic compaction which was introduced earlier to determine the level of improvement for the granular soils. Current GA program concentrates on the maximum depth of improvement, namely depth of low degree of improvement. The GA optimization technique was used to search for the dynamic compaction design which has the maximum output namely depth of low degree of improvement.
Each individual in generation contains four dependent
Case study (Asalooyeh site, Iran, 2005)
The presented method is applied to model DC performed in Asalooyeh Site, South of Iran. The soil deposit consists of 8 m loose sand overlying a layer of silty clay between 8 and 12 m depth. A 20 cm thick rock is at 4 m depth. The water table level is about 5 m below the ground surface. The standard penetration test results for this site are shown in Fig. 14. Three passes of DC using a 15.5 ton and 1 m circular tamper were carried out over a 4.6 by 4.6 m grid pattern. In the first pass, the tamper was
Conclusion
This paper has proposed a fuzzy system to determine the optimal design of soil dynamic compaction using genetic algorithm and fuzzy system analysis. The aim has been to obtain the most appropriate agreement with the field database. This presented system has six input parameters which are tamper weight, height of tamper drop, print spacing, tamper radius, number of impact, and soil layer cone resistance (qc). The outputs involve the degree and depth of improvement.
The main advantage of a fuzzy
References (26)
- et al.
An experiment in linguistic synthesis with a fuzzy logic controller
Int J Man-Mach Stud
(1975) - et al.
Theoretical and practical aspects of dynamic consolidation
Geotechnique
(1975) - et al.
Dynamic compaction of grandular soil
J Geotech Eng ASCE
(1980) - Lukas RT. Dynamic compaction for highway constructions. Washington, DC, 1986, Vol. 1, Report no....
- Lukas RT. Dynamic compaction geotechnical engineering circular. 1995, No. 1, Publication no....
- et al.
Ground response to dynamic compaction
J Geotech Eng, ASCE
(1984) - et al.
Design of dynamic compaction
Can Geotech J
(1992) - et al.
Impact of weight falling onto the ground
J Geotech Eng, ASCE
(1994) - et al.
Dynamic compaction analysis
J Geotech Eng, ASCE
(1992) - et al.
Dynamic compaction of loose granular soils eeffect of print spacing
J Geotech Eng, ASCE
(1994)
Genetic algorithms and engineering design
Cited by (28)
Performance evaluation of high energy dynamic compaction on soil-rock mixture geomaterials based on field test
2023, Case Studies in Construction MaterialsCitation Excerpt :Dynamic compaction (DC) is a versatile soil densification method for ground treatment and is extensively used due to its cost-effectiveness, timesaving, simplicity, and considerable depth of impact [7–13]. DC plays an important role in improving the mechanical behavior of loose granular materials, especially in mountainous areas with various soil types, including sands, silts, gravels, red-sandstone, and loess [6,8,9,14,15–18]. Nevertheless, it is difficult to determine some key operational parameters for performing DC because the effectiveness can be affected by:1) ground properties, including soil types, groundwater table, and underlain compressible layer; 2) DC technique, including shape, mass and drop height of the hammer, number of drops and passes, and grid spacing between impact points [7,19].
Dynamic optimization of compaction process for rockfill materials
2020, Automation in ConstructionCitation Excerpt :This is particularly suitable for multistage decision processes. Owing to its advantages, the GA has been widely used for various engineering optimization problems [29–31]. The fitness function, constraints, and algorithm parameters of GA will be introduced in Section 3.2.
Dynamic compaction of a thick soil-stone fill: Dynamic response and strengthening mechanisms
2020, Soil Dynamics and Earthquake EngineeringSimulation of dynamic compaction and analysis of its efficiency with the material point method
2019, Computers and GeotechnicsCitation Excerpt :Thus far, studies on dynamic compaction are mainly based on experiments and numerical simulations. For most of the numerical simulations, finite element methods are used [3–11], and a few researchers adopted other methods, such as the smoothed particle hydrodynamics method (SPH) [12], fuzzy theory and genetic algorithms [13], and the discrete element method (DEM) [14]. When the traditional finite element method is applied to large deformation problems such as dynamic compaction, special computational techniques of re-meshing and reassignment of material properties are usually required [15,16].
Estimation of nonlinear static skin friction on multiple pile segments using the measured hammer impact response at the top and bottom of the pile
2012, Computers and GeotechnicsCitation Excerpt :To develop a robust solution strategy for the skin friction in the finite difference equations describing the pile–soil segment system, a global inversion technique was selected because of the nonlinear nature of the inversion problem and the potential lack of information on soil/rock profiles as well as the inherent noise (i.e., potential local minima) in the measured data. Unlike local inversion techniques (e.g., methods using the gradient) that strongly depend on initial models, the global inversion techniques, e.g., simulated annealing [23,24,27] and genetic algorithms [1,8,17,18,21,22,28], which search over a larger parameter space because of their stochastic nature, use more global information to update the current position and thus are likely to converge to the global minimum. In this study, a genetic algorithm is employed because it can be used in cases in which the model-data relationship is highly nonlinear and produces multimodal misfit functions, and it is typically faster than simulated annealing [22].
A new method for estimating driven pile static skin friction with instrumentation at the top and bottom of the pile
2011, Soil Dynamics and Earthquake EngineeringCitation Excerpt :The use of any local inversion techniques (e.g., gradient methods) was ruled out because of their heavy dependence on an initial model or prior information, which are not always available real time. Also, unlike local inversion techniques, global inversion techniques, e.g. simulated annealing [24,30] and genetic algorithm [5,6,8,14,20,22,23], search over a larger parameter space due to their stochastic nature when finding the global minimum of the misfit function. In this study, a genetic algorithm is employed because it can be used in cases where the model-data relationship is highly non-linear and produces multimodal misfit functions [21], and it is typically faster than simulated annealing [22].