Houcine Hassan

The pharmaceutical industry is facing significant economic challenges due to measures aimed at containing healthcare costs and evolving healthcare regulations. In this context, pharmaceutical laboratories seek to extend the lifespan of their machinery, particularly fluid bed dryers, which play a crucial role in the drug production process. Older fluid bed dryers, lacking advanced sensors for real-time temperature optimization, rely on fixed-time deterministic approaches controlled by operators. To address these limitations, a groundbreaking approach taking into account Exploration Data Analysis (EDA) and a Catboost machine-learning model is presented. This research aims to analyze and enhance a drug production process on a large scale, showcasing how AI algorithms can revolutionize the manufacturing industry. The Catboost model effectively reduces preheating phase time, resulting in significant energy savings. By continuously monitoring critical parameters, a paradigm shift from the conventional fixed-time models is achieved. It has been shown that the model is able to predict on average a reduction of 50.45% of the preheating process duration and up to 59.68% in some cases. Likewise, the energy consumption of the fluid bed dryer for the preheating process could be reduced on average by 50.48% and up to 59.76%, which would result on average in around 3.120 kWh energy consumption savings per year.

In this paper, a data preprocessing methodology, EDA (Exploratory Data Analysis), is used for performing an exploration of the data captured from the sensors of a fluid bed dryer to reduce the energy consumption during the preheating phase. The objective of this process is the extraction of liquids such as water through the injection of dry and hot air. The time taken to dry a pharmaceutical product is typically uniform, independent of the product weight (Kg) or the type of product. However, the time it takes to heat up the equipment before drying can vary depending on different factors, such as the skill level of the person operating the machine. EDA (Exploratory Data Analysis) is a method of evaluating or comprehending sensor data to derive insights and key characteristics. EDA is a critical component of any data science or machine learning process. The exploration and analysis of the sensor data from experimental trials has facilitated the identification of an optimal configuration, with an average reduction in preheating time of one hour. For each processed batch of 150 kg in the fluid bed dryer, this translates into an energy saving of around 18.5 kWh, giving an annual energy saving of over 3.700 kWh.

Background: In technical subjects, the “hands-on” aspect is important because the students cannot assimilate some concepts until they practice with real industrial systems. They need to have the demonstration that what they are performing is real and works in order to create a true mental image of the subject concepts, which translates as an increase in professional skills. Application Design: In this work, it is described a step forward in integrating mobile devices and real physical laboratories in a ubiquitous learning approach. Contribution: The latest experiences with mobile devices showed that it is possible to get near real hands-on by combining specially designed equipment and flexible virtual-scale models running on the mobile device, without requiring physical laboratories or classic remote labs. This approach can be applied to many technical subjects that face the same issues. Findings: Based on the evaluation results, the proposed u-lab architecture provided strong evidence of the benefits achieved. The proposal presents interesting advantages from the economic and professor productivity point of view, obtaining better academic achievements. The resulting platform should facilitate educational activities which use mobile devices. This approach can be applied to many technical subjects that suffer economic restrictions. This opens up new scenarios for mobile-device-based learning without excessive additional costs.