Sensor-based quality inspection of secondary resources: Laser-induced breakdown spectroscopy

Abstract

The quality control of materials from waste is an important step for acceptance in new high-quality products. Given the large volumes and complexity of waste streams, this requires advanced sensor technology such as the Laser Induced Breakdown Spectroscopy (LIBS) investigated in this thesis. Recyclers often rely solely on manual sorting and visual inspection, which is far less reliable and accurate than sensor technology that enables automated inspection. The main source of waste in this thesis is demolition concrete. It is characterized by large amounts of moisty granular material, varying amounts and types of impurities, and a dusty environment. This presents a major challenge to introduce an optical inspection technique such as LIBS. The research starts with a literature study into the state of the art of LIBS as a material identification technique. This is followed by a more physically oriented literature review of the transient processes and parameters involved in ablation and plasma formation by a high power pulsed Nd:YAG 1064 nm laser. From this follows the development of models to account for the different processes and phases of the matter. To this end, the different transient processes are separated and analysed by making use of local equilibrium conditions. Subsequently, the influence of the optics and associated hardware that is necessary to obtain good optical data is investigated. This knowledge is supplemented by a study of light collection and data processing techniques to further improve the quality and reproducibility of the optical data, given the challenging conditions in a recycling factory. Subsequently, to better understand the potential and limitations of experimental LIBS data, the useful information about the chemical composition and technical properties of the material sample is tested under different conditions. The aforementioned separate physical models have been compiled into a complete plasma model (MLIBS for short). This can explain the properties and emissions of a laser induced plasma. This model can also be used in reverse to determine the composition of sampled elements in a laser induced plasma. The focus is on the ability not to average data to eliminate noise, but to use the data from each laser shot usefully for identification. This is called single-shot data and it is the most efficient way to implement LIBS in practice, because the price of LIBS hardware increases rapidly with the number of shots a laser has to deliver per second. The last part of research is on statistical techniques that can use all relevant information in LIBS data to make the best decision when it comes to identifying different complex materials. By accumulating statistics, the material composition of contaminated concrete waste flows can be determined more reliably. Based on the entire research in this thesis, a prototype LIBS platform has been developed and integrated with a conveyor belt for the inspection of demolition concrete in a recycling plant. The platform withstood harsh outdoor conditions and successfully demonstrated the capabilities of automated inspection with LIBS. Development of the prototype included the design and integration of a laser system, optics, electronic control equipment, real-time data acquisition system, spectral data processing software, and a weatherproof protective frame.