(257c) Carbon — I-Polypropylene Nanocomposites Synthesized with Zirconocene/MAO Catalyst System: Property Estimation through Simulation of De Assisted Dendritic Neuron Model

Industrial and academic research has shown a huge interest in polyolefin nanocomposites in recent years. These nanocomposites are high potential materials with much improved mechanical properties sch as strength, toughness, stiffness and flame retardancy.

Single site catalysts, with the common name metallocenes, are increasingly being studied and used to produce various terminal-olefinic polymers with more specific molecular architecture than those can be synthesized by conventional polymerization methods like using Zeigler-Natta type catalysts. Metallocenes, being single site catalysts, permit the synthesis of polymer nanocomposites with tailored microstructural properties and control over tactility and stereoregularity. Generally, the nano-dimensional component makes the reinforcement phase in polymer nanocomposites. These reinforced nanocomposites show a radical enhancement to the physical and chemical properties to the native polymer matrix.

The use of dendritic neural networks (NNs) has become gradually prevalent for applications where the automatous portrayal of the interrelation of dependent and independent variables is either incomprehensible or very complex.^[1] The neural network consists of processing neurons and information flow channels between the neurons, usually called ‘interconnects’. Each processing neuron calculates the weighted sum of all interconnected signals from the previous layer plus a bias term and then generates an output through its activation transfer function. Training is done by assigning random weights to each neuron, evaluating the output of the network and calculating the error between the output of the network and the known results by means of an error or objective function. If the error is large, the weights are adjusted, and the process goes back to evaluate the output of the network. This cycle is repeated till the error is small or a stop criterion is satisfied.^[2] While training an artificial NN, one needs to find out optimal weight and bias set of the network. Currently, many traditional optimization techniques, such as Backpropagation algorithm, Quasi-Newton algorithm, Genetic algorithm, Simulated Annealing algorithm, Particle Swarm Optimization algorithm etc. are widely used to train the neural networks. Differential evolution (DE), is one among the non-traditional optimization technique and it has proven, through benchmark tests, to be superior to many algorithms in terms of speed and convergence.^[3]

DE is a population based, stochastic, evolutionary algorithm, capable of finding global optima for complex optimization problems that are nondifferentiable, non-linear, discrete or multi-modal in nature, with only a few parameters.

This work deals with differential evolution assisted artificial dendritic neuron modeling for property estimation of carbon nanocomposites of iso-polypropylene synthesized with Me₂Si(2—Me-Ind₂)ZrCl₂/MAO catalyst system. The effect of temperature, monomer concentration and catalyst concentration with different fillers on catalyst activity, molecular weights and melting point of the carbon nanocomposites of iso-polypropylene synthesized with zirconocene catalyst has been investigated in MATLAB^® (R2020a) software.^[4]

References

[1] Fernandes F. A. N. and Lona L. M.F., Brazilian Journal of Chemical Engineering, 22: 03, p. 401-418 (2005).

[2] Haykin, S., Neural Networks: A Comprehensive Foundation, Prentice Hall, New York (1998).

[3] Vesterstrom, J. & Thomsen, René., Congress on Evolutionary Computation, 980-987 (2004).

[4] MATLAB. version (R2020a). Natick, Massachusetts: The MathWorks Inc. (2020).