Bauxite is a residual soil resulting from intense tropical weathering of rocks rich in aluminium oxide. It is the most important source of aluminium and is recently tapped as potential source of rare earth elements (REEs). The Philippines holds substantial bauxite deposits on Samar Island, which were declared as bauxite mining reservation sites by Presidential Proclamation in 1977. Of these reservation sites, the Paranas bauxite deposit being developed by Alumina Mining Philippines Inc., is in the most advanced stage of exploration, with numerous test pit and drill core data available. This study aims to describe the mineralogical and geochemical characteristics of the Paranas bauxites.

The bauxites in Paranas are recognized by their orange-brown clayey soil, with minor disseminations coarse black manganese particles. X-ray diffraction analysis have shown the major mineral components to be gibbsite, boehmite, goethite, lepidocrocite, cristobalite, and anatase. This study identified two types of bauxite within the area: lateritic bauxite and karst bauxite. The former arising from in situ weathering of mafic sedimentary rocks, and the latter originating from the weathering of basalts and their subsequent deposition in limestone cavities in karst landscapes. The karst bauxites, found in the southwestern area, contain higher aluminium content (>35 wt %) than the lateritic bauxites located in the northwest. The Paranas bauxites are characterized by high levels of Fe2O3, SiO2, and P2O5. Furthermore, REE concentrations reach up to 700 ppm in the karst bauxites. The ongoing study on mineral chemistry intends to map out the elemental distribution across different soil and mineral phases within these deposits, further advancing our knowledge of this essential geological resource.

The use of hyperspectral sensors has become increasingly popular for mapping ground targets. Due to the large number of spectral bands present in these sensors and the narrow spectral bandwidth, mineralogical identification and quantification are possible with precision according to the spectral signature of each mineral through a known and specific pattern of reflection and absorption at different wavelengths. To evaluate the potential use of these sensors in bauxite mining areas, mining fronts were imaged in Paragominas, in the northeast region of the State of Pará, Brazil, using the HyspeX hyperspectral sensor (SWIR 970–2500 nm). 72 samples were collected for scanning and construction of the spectral library with the assistance of a FieldSpec 3 Jr. spectroradiometer that captures 2151 bands between 350 and 2500 nm (vis-NIR). The chemical results were subjected to statistical and discriminant analysis using SAS software, and the images were processed in CaliGeo PRO and ENVI Classic software. Algorithms developed in the Python language (libraries available in Scikit Learn, MatLab, and Orange software) were used to evaluate, model, and predict the spectral curves with the chemical contents of the samples. The results showed a moderate distinction between the layers evaluated by the reflectance spectrum in the laboratory, both in shape and intensity. The use of neural networks for prediction presented the best results (60 to 90% accuracy), and the spectral range of 1000–2500 nm established more robust models for the analyzed data, attesting to the potential for application of the hyperspectral methodology for characterization and quality control of bauxite mining. However, a larger set of samples and a more robust spectral library would allow refinement and improvement of the prediction model’s performance, reducing the areas of overlap in the spectral responses.

Efficient mining operations and ore processing are vital for maximizing productivity in the mining industry. However, the inherent spatial variability in ore characteristics presents significant challenges, including material rehandling, excessive stockpiling, suboptimal plant throughput, and plant downtime. These challenges necessitate a proactive approach to ore control, as issues observed in processing and mining operations are often linked to ore characteristics. Presently, the lack of real-time forecasting of orebody characteristics entering processing plants hampers proactive adjustments, exacerbating operational inefficiencies.

In this paper we introduce the Closed-loop Real-time Ore Control System (CROCS), an innovative solution designed to revolutionize ore control in the mining industry. CROCS includes an orebody-to-product stockpile tracking system, contextualization of tracking data, and feedforward control models for predicting and optimizing mining operations and plant parameters. By seamlessly integrating pit and plant data, CROCS establishes a comprehensive ore-tracking system that monitors ore movement from extraction to processing in real-time. This enables quick identification of problematic ore and facilitates timely adjustments to operational parameters, thus mitigating material handling issues, disruptions, and suboptimal throughput. CROCS also offers opportunities for real-time reconciliation against existing plans and enhances short-term planning capabilities for material delivery strategies, ultimately optimizing resource utilization and reducing operational costs. This study presents the application of CROCS at Rio Tinto’s Amrun bauxite operation in Far North Queensland, Australia

Bauxite is the main ore for metallic aluminum production, consisting of aluminum, iron oxides and kaolinite, a clay mineral commonly found in Amazonian bauxites, as the main carrier of reactive silica. In the process, due to the small particle size, kaolinite is usually removed by attrition and washing of coarse material followed by desliming using hydrocyclones. Kaolinite has a special relevance in this context as in the Bayer process, it reacts with sodium hydroxide, increasing reagent consumption in the process. Beneficiation process at Hydro Paragominas is based on the separation of coarser fractions with higher gibbsite content from the clay minerals, where kaolinite is more concentrated. The separation takes place in two-stage hydrocyclone circuits, one for mid-size particles classification and another for fine particles classification. Pilot tests and process simulations have been carried out seeking the optimization of cyclone geometry, by testing different apex and vortex diameters. A total of 36 conditions were evaluated with optimized condition bringing reactive silica reduction potential of 1.3 percentual points.