Predicting the mechanical properties of date palm wood fibre-recycled low density polyethylene composite using artificial neural network

Due to these enumerated reasons, organic fillers are increasingly being considered in composite production. Hence, date palm wood was used as filler to develop the polymer composite used in this work. Mechanical properties of materials are extremely important and determine the quality of the material with respect to its application to engineering. Polymer composites are a very important class of engineering materials.

Nigeria is in the tropics and borders the Sahara Desert. Date palms are abundant in Northern Nigeria. Large quantities of date palm nuts are consumed in Nigeria either directly as food or indirectly as raw material in food and agro-based industries (AbdulQadir et al. 2011). Information on date palm production levels in nine states in Nigeria reveals that date palm production is increasing. The states are Bauchi 6,000 metric tons, Adamawa 600 metric tons, Borno 1,000 metric tons, Gombe 1,500 metric tons, Jigawa 5,000 metric tons, Kano 6,000 metric tons, Plateau 100 metric tons, Taraba 200 metric tons and Yobe 2000 metric tons (Nigerian Institute for Oil Palm Research 2008). As the level of production of date palm increases, the waste from it increases as well. The conversion of the waste to useful material will whittle down the impact of the waste to the environment. Again, the geographic location of date palm may influence the property of the resultant composite. In this work, particle size and particle loading were varied since mechanical properties depend strongly on them (Shao-Yun et al. 2008).

Secondly, Nigeria consumes a lot of bottled and sachet water. Most of the bottles and sachets are made with low-density polyethylene (LDPE), and these sachets constitute a major source of pollution in Nigerian cities and towns (Atuanya 2013), thanks to the volume of waste from recycled low-density polyethylene sachets used for packaging potable water generated in Nigeria which rose up to 22.4 million tons in 2012 (Akinro et al. 2012). Our choice of recycled LDPE is borne out of environmental consideration, as LDPE from sachets disposed from consumed water sachets constitutes environmental hazards in our towns, cities and villages.

The literature is replete with works done on the use of date palm fibre as fillers for composites. For instance, Sbiai et al. (2008) used epoxy as the matrix in a composite developed with date palm fibres and investigated the effect of fibre sizes on the physicochemical and mechanical properties (flexural and tensile). They reported improved mechanical properties in both the linear and non-linear range upon addition of date palm tree fibres, especially when the shorter fibres were used. Al-Kaabi et al. (2005) investigated the potential of natural fibre extracted from the date palm tree as reinforcement for polyester matrix composites. Their results show that these fibres may yield reasonable properties and could be used for low-cost applications that require low to medium strength. Mahdavi et al. (2010) compared the mechanical properties of composites made from various fibres extracted from date palm and high-density polyethylene. Bendahou et al. (2008) used compatibilized low-density polyethylene and polypropylene as matrices for a data palm fibre-based composite. They characterized morphology and thermal and mechanical properties of the resulting composites using SEM, DSC and tensile tests. They reported that compatibilization enhanced the mechanical performances for both sets of composites up to a critical amount of compatibilizer, beyond which the degree of crystallinity of the matrix decreases. Other works where date palm fibre is used as fillers for polymer composites include Wazzan (2005), Al-Khanbashi et al. (2005), etc. From the literature, it is apparent that percentage composition, particle sizes and type of fillers affect the mechanical properties of composites. One of the ways this work differs from the others mentioned above is that we used recycled low-density polyethylene for our study and investigated the effect of fibre sizes and percentage composition on six mechanical properties of the composite.

Traditionally, laboratory trial mixes have been used to determine the mechanical properties of polymer composites, as well as other engineering materials. Experimental determination of mechanical properties of materials is costly and time-consuming. Finding a potable low-cost way of predicting the strength of polymer composite materials would help in solving the problem (Nwobi-Okoye et al. 2013; Umeonyiagu and Nwobi-Okoye 2013). One way this could be done is by developing a computational model based on artificial neural network technology for predicting the strength of polymer composite materials (Nwobi-Okoye et al. 2013; Umeonyiagu and Nwobi-Okoye 2013). With mathematical and computational models, a designer can easily find the best combination of constituent materials to balance strength and cost.

An artificial neural network (ANN), usually called ‘neural network’ (NN), is a mathematical model or computational model that is inspired by the structure and/or functional aspects of biological neural networks (Russell and Norvig 2003). The concept of artificial neurons was first introduced in 1943 (McCulloch and Pitts 1943). Russell and Norvig (2003) stated that since 1943 when McCulloch and Pitts introduced the concept of neurons, much more detailed and realistic models have been developed both for neurons and for larger systems in the brain, leading to the modern field of computational neuroscience. Since the work of McCulloch and Pitts in 1943, ANN has had wide application in many spheres of life. According to Maier and Dandy (2000), in recent years, ANNs have become extremely popular for prediction and forecasting in a number of areas, including finance, power generation, medicine, water resources and environmental science.

The utility of artificial neural network models lies in the fact that they can be used to infer a function from observations. This is particularly useful in applications where the complexity of the data or task makes the design of such a function by hand impractical (Russell and Norvig 2003).

The literature is replete with works on the application of ANNs for the prediction of properties of engineering materials. Ozerdem and Kolukisa (2009) used ANN to predict the mechanical properties of Cu-Sn-Pb-Zn-Ni cast alloys. In the work, Cu-Sn-Pb-Zn-Ni weight percent (wt%) contents were used as inputs, while yield strength, tensile strength and elongation were the outputs. The results of the study showed that neural network was successful in the prediction of yield strength, tensile strength and elongation of Cu-Sn-Pb-Zn-Ni cast alloys. Hassan et al. (2009) used ANN in predicting some physical properties of (porosity and density) and hardness of aluminium-copper/silicon carbide composites. The result they obtained showed that maximum relative error for predicted values did not exceed 5.99%. Bilim et al. (2009) used ANN to predict the compressive strength of ground granulated blast furnace slag concrete. The results showed that ANN can be an alternative approach for predicting the compressive strength of the material. Asilturk and Cunkas (2011) used artificial neural network and multiple regression method to predict the surface roughness in turning operations. The results show that ANN predicted better than the models based on regression. Mukherjee and Biswas (1997) in their paper applied artificial neural networks for the prediction of the mechanical behaviour of concrete materials at high temperature. Their results were very encouraging. Oretal and Kawashima (2003) in their paper proposed an ANN-based model to predict the confined compressive strength and corresponding strain of circular concrete columns. Their study shows the importance of validating the ANN models in simulating physical processes especially when data are limited. The ANN model they developed was also compared to some analytical models and was found to perform well. Other papers on the prediction of material strength using neural networks include the following: Lee (2003), Kasperkiewicz et al. (1995), Ahmet et al. (2006), Topcu and Saridemir (2008), etc.

Maier and Dandy (2000) reviewed 43 papers dealing with the use of neural network models for the prediction and forecasting of water resource variables in terms of the modelling process adopted. They identified inadequate model building as the obstacle militating against accurate predictions using artificial networks. They suggested that ANN models must be properly evaluated before their application in time series analysis.

Their assertion is corroborated by Chatfield (1993) when commenting on the suitability of ANNs for time series analysis and forecasting, who commented thus ‘when the dust has settled, it is usually found that the new technique is neither a miraculous cure-all nor a complete disaster, but rather an addition to the analyst's toolkit which works well in some situations and not in others’.

It is important to note that a neural network modelling is purely a computational technique. Hence, if one wants to explain an underlying process or mathematical framework that produces the relationships between the dependent and independent variables, it would be better to use a more traditional statistical model like regression analysis. However, if model interpretability is not important, one can often obtain good model results more quickly using a neural network.

Reinforcement is generally responsible for strengthening the composite and improves its mechanical properties. All of the different fibres used in composites have different properties and so affect the properties of the composite in different ways. This also provides stiffness to the composites (Myers et al. 1991).

The advantages of composite materials over conventional materials stem largely from their higher specific strength, stiffness and fatigue characteristics, which enables structural design to be more versatile. They also have good resistance to corrosion, low cost, increased toughness and good resistance to fire (Bledzki and Gassan 1999).

Fillers find application in the polymer industry, almost exclusively to improve mechanical, thermal and electrical properties and dimensional stability (George et al. 2001). Fillers increase the number of chains, which share the load of a broken polymer chain.

Neural network approach was used to predict the mechanical properties of date palm wood fibre-low density polyethylene composite. Predicting mechanical properties of polymer composite is necessary in engineering design applications (Mukherjee and Biswas 1997; Lee 2003). The neural network model developed has intuitive and theoretical appeal. It was developed based on the assumption that the experimental results were generated by a stochastic process. The model developed was in very good agreement with values obtained from the experiment.

Summarizing, in this work, low-density polyethylene from containers used to bag potable water was gathered, recycled and combined with date palm wood fibre flour to produce a composite material. The mechanical properties of the material, namely the ultimate tensile strength, elongation, tensile modulus, flexural strength, flexural modulus and hardness, were experimentally determined for various percentage weight compositions of date palm wood fibre ranging from 0% to 30% and for various particle sizes, namely 150, 212, 250 and 300 μm. Later, the experimentally determined mechanical properties of the composite at various percentage compositions of the date palm wood fibre and at various particle sizes were used to train an ANN; the trained ANN was later used to predict the mechanical properties of the composite at various percentage compositions of the date palm wood fibre and at various particle sizes.