Browsing by Author "Vrancken, Carlos"

Now showing 1 - 5 of 5
  • Thumbnail Image
    ItemOpen Access
    Critical review of real-time methods for solid waste characterisation: Informing material recovery and fuel production
    (Elsevier, 2017-01-28) Vrancken, Carlos; Longhurst, Philip J.; Wagland, Stuart Thomas
    Waste management processes generally represent a significant loss of material, energy and economic resources, so legislation and financial incentives are being implemented to improve the recovery of these valuable resources whilst reducing contamination levels. Material recovery and waste derived fuels are potentially valuable options being pursued by industry, using mechanical and biological processes incorporating sensor and sorting technologies developed and optimised for recycling plants. In its current state, waste management presents similarities to other industries that could improve their efficiencies using process analytical technology tools. Existing sensor technologies could be used to measure critical waste characteristics, providing data required by existing legislation, potentially aiding waste treatment processes and assisting stakeholders in decision making. Optical technologies offer the most flexible solution to gather real-time information applicable to each of the waste mechanical and biological treatment processes used by industry. In particular, combinations of optical sensors in the visible and the near-infrared range from 800 nm to 2500 nm of the spectrum, and different mathematical techniques, are able to provide material information and fuel properties with typical performance levels between 80% and 90%. These sensors not only could be used to aid waste processes, but to provide most waste quality indicators required by existing legislation, whilst offering better tools to the stakeholders.
  • Thumbnail Image
    ItemOpen Access
    Deep learning in material recovery: Development of method to create training database
    (Elsevier, 2019-02-05) Vrancken, Carlos; Longhurst, Phil; Wagland, Stuart
    Increasing the rate of material identification, separation and recovery is a priority in resource management and recovery, and rapid, low cost imaging and interpretation is key. This study uses different combinations of cameras, illuminations and data augmentation techniques to create databases of images to train deep neural networks for the recognition of fibre materials. Using a limited set of 24 material samples sized 1200 cm2, it compares the outcome of reducing them to 30 cm2. The best classification accuracies obtained range from 76.6% to 77.5% indicating it is possible to overcome problems such as limited available materials, time, or storage capabilities, by using a setup with 5 cameras, 5 lights and applying simple software image manipulation techniques. The same method can be used to create deep neural network training databases to recognise a wider range of materials typically found in solid waste streams, in real-time. Furthermore, it offers flexibility as the classification cameras could be deployed at different stages within solid waste processing plants, providing feedback for process control, with the potential of increasing plant efficiency and reducing costs.
  • Thumbnail Image
    ItemOpen Access
    Development of a method to classify and analyse the composition of mixed waste materials in real-time.
    (Cranfield University, 2022-07) Vrancken, Carlos; Wagland, Stuart; Longhurst, Philip
    There is a need for innovative technologies to classify and monitor the composition of solid waste in real-time. This research project has highlighted which information is required to improve current process designs. It also identified visible spectrum cameras as the solution that can better inform waste composition and quality without requiring complementing technologies. The experiments applied deep learning methods to classify the materials based on their images, and a method to analyse the composition of mixed waste was developed. There is a high variability in the appearance of waste materials in the context of a material recovery facility. An image capture setup using multiple cameras and light sources was implemented and tested to acquire a representative set of images. The hardware captures images from different angles, with enhanced shadow details, and different exposure levels. Image processing software further augmented the data by rotating and changing the images resolutions. The images were converted to greyscale to increase the method robustness without affecting classification performance. Deep convolutional neural networks were trained on the augmented datasets. The trained networks obtained state-of-the-art performance when tested and validated for the task of waste material classification. Based on this, a composition analysis methodology was developed and tested with mixed material samples. The methodology provides results as accurate as current manual solutions, but it can analyse a waste stream on a conveyor belt in real-time. The findings and observations from the experimental results contribute to knowledge in three main areas: data capture, data processing, and deep learning. This thesis presents the progressive development of the methodology and discusses different applications for waste management. The composition analysis can provide real-time waste data to improve the overall efficiency of the waste treatment industry. This information can be also used by stakeholders for better decision-making in the future.
  • No Thumbnail Available
    ItemOpen Access
    Results from deep learning tests using balanced databases for the classification of paper and cardboard materials.
    (Cranfield University, 2019-10-14 14:28) Vrancken, Carlos; Wagland, Stuart; Longhurst, Philip
    For methodology used to obtain these results please refer to the publication: "Deep learning in material recovery: Development of method to create training database".These results were obtained using grayscale version of the images.The "Balanced dataset - classification results" spreadsheet includes:Sheet 1 - classification results when classifying 3 classes of fibre materials using increasing number of samples per class in a balanced training datasetSheet 2 - classification results when using a balanced dataset with 5,000 training samples per class to classify 10 classes of fibre waste material
  • Thumbnail Image
    ItemOpen Access
    Scaling Milton Keynes power requirements for electrical transportation
    (Institute of Electrical and Electronics Engineers, 2016-07-18) Nieto-Martin, Jesus; Butans, Eugene; Correa-Castillo, Pablo; Fontaine, Jerome; Gonnelle, Alexis; Lahjibi, Mohamed; Vrancken, Carlos
    Milton Keynes is home to the UK’s first installation of a wirelessly charged passenger bus route. This Inductive Power Transfer (IPT) system enables a fleet of 8 electric buses to service a demanding 15-mile urban route. Opportunistic wireless charging of the batteries during the layover time at the routes allows reducing the size of the batteries, consequently improving cost and performance characteristics of the bus. This paper aims to analyze the effects of electric buses on the electricity distribution grid. In particular, the paper analyses scalability of the IPT solution to all urban routes in Milton Keynes and compares peak power requirements generated at different points in the network with typical industrial and commercial (I&C) loads.