Content area
The sustained growth of the pharmaceutical industry, combined with the increasing complexity of supply chains, has intensified the demand for more efficient, reliable and technologically integrated logistics solutions. With increasing demands for traceability, safety and speed in the distribution of healthcare products, there is a need to think carefully about storage and order preparation systems, especially in contexts with high turnover, such as automated logistics centres.
In this context, A-Frame dispensers, widely adopted for the automatic picking of high-turnover items, depend critically on the efficiency of replenishment processes to maintain optimum performance levels. Carrying out this process manually, without digital support and systematic prioritisation criteria, leads to vulnerabilities with a direct impact on the system’s operational reliability and efficiency.
This dissertation was carried out at Medlog, a company belonging to the Cooprofar group, with the objective of analysing the refuelling process of A-Frame systems in a real-life context. The goal was to identify operational limitations and propose solutions geared towards greater control, automation and efficiency. The project began with a diagnosis of operations, based on direct observation, analysis of historical data, interviews with operators and ergonomic assessment of workstations. The results of this diagnosis revealed a frequency of logistical failures, attributed to the absence of product in the channels when orders were being prepared, a phenomenon associated with exclusive dependence on the operators’ visual perception.
Given this scenario, three proposals were designed and analysed, differing in terms of technological complexity. The first proposal involves the installation of automatic vertical modules for storage and replenishment, allowing physical and digital integration with the warehouse management system. The second is based on the development of a predictive model, supported by automatic learning algorithms, capable of anticipating replenishment needs based on historical consumption patterns. The third proposes the application of industrial cameras equipped with computer vision technology, capable of autonomously detecting the level of stock in each channel. The solutions designed were validated in a simulated environment using a discrete event model implemented in Python using the SimPy library.
The results obtained show improvements in terms of system reliability, reduced response times and improved working conditions. It was found that the success of the implementation depends on the compatibility between the technological solution adopted and the physical and organisational reality of the company. The comparative analysis of the solutions showed that although the alternative based on automated storage provides the best overall performance, the predictive model represents a balanced intermediate solution, combining moderate investment with positive operational impacts and ease of implementation. In turn, the solution based on computer vision showed high potential, but with more demanding technical and financial requirements.