Versatilist

This case study explores the development of an innovative AI-driven Till Slip scanning solution for the Retail industry. The project aimed to bridge the gap between Fast-Moving Consumer Goods (FMCG) companies and their end consumers by enabling direct engagement through a unique loyalty program. This solution represents a world-first in the industry, leveraging cutting-edge technology to transform the traditional B2B nature of FMCG companies into a direct-to-consumer model. The core functionality of the solution allows users to scan their till slips from any participating retailer. The scanned till slips are then processed using Optical Character Recognition (OCR) and advanced Artificial Intelligence (AI) algorithms to extract and match the data to the actual product information from the FMCG company. If the till slip includes any qualifying products, the user is rewarded with loyalty points, which can be redeemed through the program. The development process was comprehensive, involving several critical phases to ensure the delivery of a robust and effective solution. It began with detailed consultations with FMCG companies and retailers to understand their needs and challenges, followed by the selection of state-of-the-art OCR technology and the development of AI algorithms tailored to parse till slip data and match it with the FMCG product database. The system design focused on creating a scalable architecture that supports real-time processing and high-volume transactions. A user-friendly mobile application was developed to facilitate till slip scanning and loyalty point management. Rigorous testing was conducted to ensure accuracy and reliability, followed by pilot testing with select retailers and FMCG products to validate the end-to-end process. By bypassing the need for integration with multiple enterprise point-of-sale (POS) systems, the solution addressed one of the significant barriers to market for multi-retailer loyalty programs. Instead, it directly interpreted till slips, eliminating the need for complex and costly POS integrations. This innovative approach allowed FMCG companies to gain direct access to consumer purchasing data, providing valuable insights into purchasing behaviours and preferences. The AI-driven Till Slip scanning solution delivered significant benefits, including enhanced customer engagement, valuable consumer insights, operational efficiency, and innovative market positioning. This project exemplifies the potential of AI and OCR technologies in transforming traditional business models and driving direct consumer engagement, providing a scalable and effective means for FMCG companies to connect with their end users and enhance their loyalty programs Read More

This case study details the design and development of an innovative Predictive Maintenance solution for large mining dump trucks using machine learning. The project, initiated as a comprehensive research and development effort in the Natural Resources sector, aimed to enhance the operational efficiency and minimise the downtime of heavy mining equipment. Mining operations are critically dependent on the availability and performance of large dump trucks for transporting extracted materials. Unplanned breakdowns of these trucks can lead to significant operational delays and increased costs. The primary objective of this project was to create a predictive maintenance system capable of anticipating potential failures and scheduling maintenance proactively, thereby reducing downtime and improving overall productivity. The development process involved several key phases. Initially, large mining dump trucks were equipped with a range of sensors to monitor various parameters, including engine temperature, oil pressure, vibration levels, hydraulic system performance, and tire conditions. Historical maintenance records, operational logs, and failure incidents were also collected to provide a comprehensive dataset for model training. Data preprocessing was a crucial step in this process, involving the cleaning of raw sensor data to remove noise and outliers, handling missing values, and extracting relevant features through feature engineering. Maintenance records and failure logs were used to label the data, identifying instances of component failures and maintenance events. Several machine learning algorithms, including decision trees, random forests, gradient boosting machines, and deep learning models, were evaluated during the model development phase. The dataset was split into training and validation sets, and cross-validation techniques were employed to ensure model robustness. Hyperparameter tuning and optimisation further improved predictive accuracy and performance. Anomaly detection algorithms were also implemented to identify unusual patterns and deviations from normal operating conditions, providing early warning signs of potential failures. The predictive models were deployed on edge computing devices installed on the dump trucks, enabling real-time data processing and decision-making at the source. A cloud-based infrastructure was established to collect, store, and analyse data from multiple dump trucks across different mining sites, facilitating centralised monitoring and continuous model updates. A user-friendly interface was developed for maintenance personnel, providing real-time alerts, diagnostic reports, and maintenance recommendations through interactive dashboards. The solution was initially deployed in a pilot phase involving a select fleet of dump trucks at a major mining site to validate the system’s performance in real-world conditions. Following successful pilot testing, the Predictive Maintenance solution was rolled out across the entire fleet of dump trucks, accompanied by training sessions for maintenance personnel. The implementation of the Predictive Maintenance solution resulted in significant improvements in the operational efficiency of the mining dump trucks. Unplanned breakdowns were substantially reduced, leading to increased availability and minimised operational delays. Proactive maintenance scheduling helped avoid costly emergency repairs and extend the lifespan of critical components, while early detection of potential failures contributed to safer operating conditions. Additionally, the system provided valuable insights into the health and performance of the dump trucks, enabling data-driven decision-making for maintenance strategies and operational planning. This project underscores the transformative potential of machine learning in enhancing maintenance practices within the Natural Resources sector, demonstrating how advanced data analytics and real-time monitoring can deliver substantial operational benefits. Read More

This case study examines a research and development (R&D) project undertaken by a large water utility in collaboration with a specialist engineering company, aimed at enhancing the detection of leaks in large-diameter bulk water distribution pipelines. The primary objective of the project was to develop a reliable method for identifying leaks using an innovative airborne thermal remote sensing solution. The methodology employed involved a combination of advanced imaging technology and precise geographic information systems (GIS). The process was iterative, with multiple hardware and software versions developed and tested to achieve optimal performance. Initially, a basic mount for the thermal imager and standard camera was attached to the helicopter door. However, this setup resulted in excessive interference, compromising the quality of the data collected. To address this issue, the team upgraded to a custom-built mount installed on the nose of the helicopter. This upgrade included the integration of a machine vision imagery camera and a Germanium lens for the thermal camera, which significantly improved data quality. Despite these enhancements, further refinement was necessary to achieve the desired results. In the final iteration, a collaboration with a third-party aerospace company led to the design and construction of a 5-axis gyro-stabilised gimbal from the ground up. This advanced gimbal provided exceptional stability for the cameras, greatly enhancing the accuracy and reliability of the data captured. Additionally, the thermal camera was upgraded to further improve detection capabilities. To complement these hardware advancements, custom software was developed in MatLab to handle the image streams from both cameras. This software synchronised the imagery, allowing for precise side-by-side analysis. The collected data was then analysed using this proprietary software, which matched the frame rate and position of both cameras, facilitating detailed examination of potential leaks. The integration of these advanced technologies and the iterative approach to development culminated in a robust solution capable of effectively identifying leaks in bulk water distribution pipelines. The success of this project underscores the importance of interdisciplinary collaboration and the application of cutting-edge technology to address complex challenges in the utilities sector. Read More