Elsevier

Materials & Design

Volume 31, Issue 4, April 2010, Pages 1900-1905
Materials & Design

Prediction of mechanical properties of polypropylene/waste ground rubber tire powder treated by bitumen composites via uniform design and artificial neural networks

https://doi.org/10.1016/j.matdes.2009.10.057Get rights and content

Abstract

Polypropylene (PP)/waste ground rubber tire powder (WGRT) composites were studied with respect to the effect of bitumen and maleic anhydride-grafted styrene–ethylene–butylene–styrene (SEBS-g-MA) content by using the design of experiments (DOE) approach, whereby the effect of the four polymers content on the final mechanical properties were predicted. Uniform design method was especially adopted for its advantages. Optimization was done using hybrid artificial neural network–genetic algorithm (ANN–GA) technique. The results indicated that the composites showed fairly good ductibility provided that it had a relatively higher concentration of bitumen and SEBS-g-MA under the studied condition. A quantitative relationship was presented between the material concentration and the mechanical properties as a set of contour plots, which were confirmed experimentally by testing the optimum ratio.

Introduction

Our laboratory has been focusing considerable attention on development of technologies to effectively recycle waste ground rubber tire [1], [2], [3]. Recycling of waste ground rubber tire requires special techniques because waste ground rubber tire is a thermoset material, which cannot be reprocessed like thermoplastics. Powder utilization is one of the attractive techniques for effective utilization of waste ground rubber tire. A promising way of ‘recycling’ waste ground rubber tire powder (WGRT) is to incorporate it into thermoplastics to obtain thermoplastic elastomers (TPEs). In the first stage of the 10 year project (2000–2002), we developed technologies to produce ultra fine powder from waste ground rubber tire. In the second stage (2003–2005), we developed technologies to manufacture thermoplastic elastomers from waste ground rubber tire powder. In the third stage (2006–2009), we are developing technologies for mass production of TPEs based on waste ground rubber tire powder. One of major criteria for a thermoplastic elastomer is elongation at break is more than 100%. In order to achieve the target, bitumen and compatibilizer were added to the polypropylene/waste ground rubber tire powder (PP/WGRT) blend systems. According to the former papers, we know that bitumen acts as a plasticizer for polyolefin [4]. In addition, bitumen also acts as a devulcanized agent for WGRT [5], [6]. The interface adhesion between PP and WGRT is usually very weak due to the crosslinked structure of WGRT. In order to solve the problem, some attempts were made to produce thermoplastic rubbers [7], [8]. It was early recognized that WGRT should be devulcanized or at least partially devulcanized to facilitate the molecular entanglement between PP and WGRT. In order to further improve the interface adhesion between PP and WGRT, the compatibilizer and bitumen are added together.

The sharp market competition makes it important to shorten the development cycle of products and reduce the costs. The application of statistical experimental design and analysis technology in the rubber industry provides composites with a convenient, accurate and quantitative means. Statistical design of experiments (DOE) has been long used to provide efficient approaches to optimize process parameters and rubber formulary in rubber processing [9]. Most statistical experimental designs usually adopted in rubber composites are two level factorial design, screening design and response surface design, in which the response surface design especially has been used widely because of its use of less experimental trials and its capability of fitting quadratic regression equations [10]. However, with the increase of experimental factors, the number of coefficients of the quadratic equation increases exponentially and hence, does the number of experimental trials. More recently, a new statistical design of experiments, the uniform design method is found to overcome this problem. Keeping this in mind, a three factor with seven trials uniform experimental design U7 (73) was chosen and experiments were carried out.

Uniform design (UD) is a useful and simple method, which was first proposed by Fang et al. [11]. Generally speaking, UD is a form of “space filling” design for computer experiments [12]. In order to establish a UD, one needs to find suitable design points that are scattered uniformly on the experimental domain. However, if we restrict the domain to certain lattice points, then UD is also an efficient fractional factorial design. For a given measure of uniformity M, a uniform design has the smallest M-value over all fractional factorial designs with n runs and m q-level factors. Examples of successful applications of the UD method for improving technologies in various fields have been consistently reported [13], [14], [15], [16], [17].

An artificial neural network (ANN) approach is a powerful mathematical tool in recognizing and modeling of material properties. They have been used in diverse applications in control, robotics, pattern recognition, forecasting, power systems, manufacturing, optimization, signal processing, etc. The idea to solve engineering problems using neural networks was developed in the 1940s in the United States. It has been introduced into the fields of materials recently [18], [19]. They are actually a computing system containing simple processing elements known as units or nodes connected by links. Each neuron generates an output signal and this is in the case that the weighted sum of inputs is greater than an actuation value. Output of each neuron is related to the inputs through transfer or actuation functions. These subsystems are organized in a series of layers. In each neuron, the scalar input p is transmitted through a connection that multiplies its strength by the scalar weight w, to form the product wp, again a scalar. This product may be added to a scalar bias (much like a weight, except that it has a constant input of 1) to form the final argument of the transfer function. A neural network is not constructed for the solution of a specific mathematical problem and no mathematical or engineering principles are applied to determine the system output. They just apply the gained knowledge from previously solved examples to build a system of neurons that learn how to solve a new examined problem by adapting the intensity of the links between nodes [20].

On the other hand, genetic algorithm (GA) is a direct search algorithm that is based on the natural evolution concept coming from Darwin’s theory of evolution. It is actually a probability-based optimization algorithm, started with an initial population of design variables. In GA, natural selection increases the surviving capability of the populations over the foregoing generations. The characteristics of each design are used to generate a fitness value indicating its level of performance with respect to the other designs in the population. Designs that perform the best (i.e., have the highest fitness value) are given the greatest probability of breeding with other good designs so that their characteristics can be passed to future generations. ANN and GA are the most promising natural computation techniques. In recent years, ANN has become a very powerful and practical method to model very complex nonlinear systems [21], [22], [23], [24], [25] and can be found in various research fields for parameter optimization [22], [23], [24], [26].

In the present work, polypropylene (PP)/waste ground rubber tire powder (WGRT) composites were prepared. The effect of bitumen and maleic anhydride-grafted styrene–ethylene–butylene–styrene (SEBS-g-MA) of various concentrations on the mechanical properties was studied. Instead of studying the effect of each factor one at a time and finally optimizing, design of experiments (DOE) methodology was chosen and uniform design method was especially adopted involving a minimum of only seven experiments. Optimization was done using a hybrid artificial neural network (ANN)–genetic algorithm (GA) approach. The primary focus of this article is to predict the mechanical properties of PP/WGRT composites companied with bitumen and SEBS-g-MA using ANN–GA approach with minimum number of experiments. Results from the predicted properties were also experimentally confirmed.

Section snippets

Materials

Polypropylene (PP R520Y, MFI = 1.8 g/10 min) was manufactured by SK Corporation in Korea. Maleic anhydride-grafted styrene–ethylene–butylene–styrene (SEBS-g-MA, Kraton FG-1901X) were obtained from Shell Chemical Co. Ltd. Bitumen (X-4) was obtained from Waterproofbank Company in Korea. Waste ground rubber tire powder (WGRT) of about 50 meshes was produced by Hongbok Industries in Korea.

Sample preparation

Waste ground rubber tire powder (WGRT) was devulcanized in a 30 mm single-screw extruder. First WGRT was passed

Results and discussion

As observed in our preliminary studies, mixtures of PP and WGRT do not lead to the formation of composites having appreciable mechanical properties. Accordingly, it was presumed that incorporation of bitumen and SEBS-g-MA in the PP/WGRT composite could produce the desired properties. Bitumen was chosen because of its plasticized effect for PP [4] and its devulcanized effect for WGRT [5], [6]. Namely the sulfur crosslink in WGRT are broken under mechanical stress (the first extrusion) [5] and

Conclusions

PP/WGRT composites with bitumen and SEBS-g-MA of various concentrations were investigated using DOE methodology to optimize a recipe for commercial applications having high mechanical properties. A proprietary software package called Rubber Computer Aided Design (RCAD) was used for this purpose. RCAD utilizes the uniform design technique for the design of starting experiments which was selected for reducing the number of preliminary experiments when compared to traditional simultaneous methods,

Acknowledgements

The authors gratefully acknowledge financial support of this research by the Science and Technology Activities of the Returned Overseas Chinese Scholars Projects Funded Merit-based Funding of Ministry of Personnel of China (080416091811).

References (28)

  • J.K. Kim et al.

    Study on powder utilization of waste tires as a filler in rubber compounding

    Rubber Chem Technol

    (1999)
  • J.K. Kim et al.

    The biological and chemical desulfurization of crumb rubber for the rubber compounding

    J Appl Polym Sci

    (1999)
  • A.M. Shanmugharaj et al.

    Modification of rubber powder with peroxide and properties of polypropylene/rubber composites

    J Appl Polym Sci

    (2007)
  • E. Lievana et al.

    Use of ground tyre rubber (GTR) in thermoplastic polyolefin elastomer compositions

    Prog Rubber Plast Recycl Technol

    (2004)
  • Cited by (65)

    • Machine learning prediction of mechanical properties of braided-textile reinforced tubular structures

      2021, Materials and Design
      Citation Excerpt :

      The obtained results from impact testing verify the accuracy of the model. Zhang et al. [27] established a quantitative relationship between the material concentration and the mechanical property via genetic algorithms, and the optimized model was confirmed to have high accuracy. Liu et al. [28] developed a new failure criterion based on structural genome and deep learning neural network model.

    • Recycling of waste tire rubber: Microwave devulcanization and incorporation in a thermoset resin

      2017, Waste Management
      Citation Excerpt :

      Several methods have been applied to devulcanize the Waste Tire Rubber (WTR) for reuse in rubber processing. Thermomechanical shearing devulcanization (Radhesh Kumar et al., 2002; Zhang et al., 2010a, 2010b; Sutanto et al., 2006a) is one of the most employed techniques. During this process, a strong rise in temperature occurs that can lead to the degradation of the rubber network.

    View all citing articles on Scopus
    View full text