Today's world is no longer that of yesterday, the pace with which we live and also the speed is enormous and rapid, that overnight we discover the appearance of new technologies and solutions in all the fields, in particular, that of scientific research. Artificial intelligence plays the main role. Predicting the behavior of new materials using artificial neural networks has become a frequently adopted solution by researchers today. The performance of neural networks depends mainly on the activation functions used. This work was designed to mainly study the impact of these functions on the response speed of an artificial neural network in general, and particularly on the model we are working on to predict the thermomechanical behavior of innovative materials. By using TANSIG, PURELIN and LOGSIG in a feed forward back propagation training by Levenberg-Marquardt algorithm, we were able to generate 9 models. For each of these models, we were interested in analyzing the speed’s response of the network and studying its regression. Thus, this work was able to show us that choosing the right neuron activation function from one layer to another can clearly influence the performance of the results. Depending on the problem studied, the desired objective and the chosen architecture, the activation function can radically change the result and provide us with the expected efficiency.

The poultry industry is responsible for a significant amount of waste. This bio-load is generally transformed into food with low nutritional value, or else incinerated. Not only are these options unsustainable, but it is also a missed opportunity to contribute to the circular economy. In this article, we try to contribute to the development of methods and strategies necessary to transform poultry feathers from waste for disposal into valuable raw materials. The chicken feathers are collected and washed several times, after drying in an oven, and finally crushed to obtain a flour which will be a bio-filler for the PVC used in this flexible article. Different percentages are used to find the one that most improves the mechanical characteristics of recycled PVC. The tensile test is carried out on test specimens made of recycled PVC without additives and recycled PVC reinforced with 5, 10 and 15% chicken feathers. It is deduced from the test results that the value of 10% was the most adequate to improve the mechanical characteristics of recycled PVC.

Neural networks have led to the evolution of the processing methodology of computational sciences. The problems like bio composites modeling and prediction are difficult to model with classical mathematical and statistical tools because of the data inherent noise. NN’s processing capability in the forecasting, recognition, modeling, system analysis and control can give fast characterization, modeling and prediction of bio composites properties, provided as long as datasets are available. Using Matlab®, a neural network model was evaluated to characterize the optimal properties of the ANS reinforced the Polypropylene. The feed forward multilayer model provided best results in comparison with the finite element method and the experimental tensile tests. The trained neural network is able to provide a best prediction of such bio composite based on natural particles having more advantages to the environment, economy and the sustainable development.

Bio composites are a new category of materials using natural based components in their constituents. The study and simulation of the behavior of these innovative materials occupies an important place in the field of scientific research. Discovering and using new methods has always been the goal of researchers. In recent years, artificial intelligence has been very successful and is used in several fields. it represents a big part of today’s industrial revolution. Smart solutions are more and more favored over conventional solutions as they give more precise results in a short time. We can find them in different sectors, such as banking, commerce, transport and industry, especially in materials science.The intersection of the artificial intelligence with materials engineering, gives extraordinary results. This smart method was able to boost the discovery of new materials, and to solve the most complex problems encountered when determining the mechanical properties of bio composites. What characterizes theEco-composites is their light in weight, their sustainable development, and that they are environmentally friendly. However, the determination of their mechanical properties is not obvious. Certainly, solutions based on homogenization methods or even on the finite element method have given good results, but the complexity of the microstructure of these materials limits the determination of their characteristics. In our paper, we hilight the use of Deep Learning that is an artificial intelligence machine learning method that relies on neural networks to predict the mechanical behavior of a polypropylen bioloaded by the natural fibers.

Phosphate mining companies in Morocco pose many environmental and occupational safety risks, especially through the release of airborne particulates, gas pollutants, and heavy metals. While there is increased implementation of monitoring systems within industrial mining contexts, current methodologies are still predominantly founded on rule-based systems or classical statistics that presume linearity in relationships between an arbitrary set of environmental parameters and the likelihood of an incident. Conversely, mining operations are characterized by intricately dynamic nonlinear combinations of numerous environmental and operational variables. As a result, a potential research opportunity exists for the application of sophisticated machine learning techniques that provide the ability to detect various levels of operational risk within phosphate mining scenarios. This study has three objectives. First, to examine the mining environmental and operational data from the phosphate mining sites to determine the mining operational conditions that present the highest risk. Second, to create a machine learning classification model which utilizes a Feedforward Neural Network (FNN) to identify operational states that are prone to incidents based on multivariate sensor data. Third, to assess the validity and reliability of the model using machine learning validity and reliability evaluation techniques along with statistical validation methods. In this study, an artificial intelligence-based approach for AI-based safety monitoring was proposed by using a Feedforward Neural Network (FNN) on a detailed data set of 1536 hourly measurements, directly recorded onsite at OCP plants in Benguerir and Khouribga. Environmental and industrial parameters (dust concentration, gas emissions, temperature, and toxic metal content) were measured using industrial-grade sensors certified for such a type of application. By means of training the proposed FNN model with adaptive gradient descent and dropout regularization with early stopping, a test mean squared error of 0.057 and over 85% accuracy on incident detection were obtained. Gradient tracking and m-adaptive validation proved the stability and convergence of the model. Emissions and dust were identified as the main risk classifiers in a variable importance analysis. The findings demonstrate that the mining sector may move from reactive to proactive safety management and validate the incorporation of AI into a real-time monitoring infrastructure inside the OCP ecosystem. Practical concerns of industrial data gathering, model interpretability, and the moral application of AI in high-risk settings are also addressed by the study.