SPECTRAL EVOLUTION Mining Applications and Chemometrics www spectralevolution
- Slides: 27
SPECTRAL EVOLUTION Mining Applications and Chemometrics www. spectralevolution. com
SPECTRAL EVOLUTION ØIncorporated 2004 ØFull line supplier of UV-VIS-NIR spectrometers for lab, inline process & field portable remote sensing ØMfg facility in North Andover , MA ØOEM manufacturer Ø>100 field portable UV-VIS-NIR instruments in field use worldwide www. spectralevolution. com
Products offered �Field portable full range UV-VIS-NIR spectrometers & spectroradiometers �Laboratory full range UV-VIS-NIR spectrometers & spectroradiometers �Single detector In. Ga. As photodiode array lab spectrometers �Single detector Si spectrometers, spectroradiometers & spectrophotometers �Light sources & accessories
Mining Spectrometers for mining exploration, mineral identification, and production � ore. Xpress™ Full range portable spectrometer for mining and mineral identification � ore. Xpress Platinum Also includes a range of FOV lenses, internal battery, membrane control panel for standalone operation, and on-board storage for 1, 000 spectra
Field Portable units ore. Xpress & ore. Xpress Platinum � True field portability <7 lbs � Full range UV/VIS/NIR – 350 -2500 nm � Fast/High Signal to Noise ratio for better reflectance values � Unmatched stability & performance through SWIR 2 � DARWin SP Data Acquisition software saves scans as ASCII files for use with 3 rd party software � EZ-ID real-time mineral ID with USGS & Spec. MIN libraries
Real-time Mineral ID EZ-ID™ Software with Library Builder Module � Real-time mineral identification in the field � USGS and Spec. MIN libraries � Select different spectral regions of interest � Compare unknown mineral sample spectra to known library � Best match score quickly and automatically displayed
Qualitative & Quantitative Analysis � Use EZ-ID for mineral identification and qualitative analysis What is there � Use reflectance spectroscopy and chemometrics for quantitative analysis How much is there
Reflectance Spectroscopy � Widely used in mining exploration and mineral identification � Identification of key alteration minerals associated with potential economic deposits � Qualitative mineralogy describes the process of using NIR to quickly ID mineral species during exploration u. Advanced Argillic u. Phyllic u. Propylitic u. Potassic
Quantitative Mineralogy � Usage is typically bound by cost (high) and speed (slow) � Available examples: Qemscan/MLA Quantitative X-ray diffraction � Better solution – Quantitative Reflectance Spectroscopy Analyze a greater number of samples in less time, at an affordable cost
Quantitative Reflectance Spectroscopy � Use mineralogical and metallurgical information from a representative set of samples and correlated reflectance spectra to develop statistical calibration models � Calibration “trains” the spectrometer to analyze additional unknown samples � Leverage the detailed, more costly analysis of a few samples to analyze a much larger set of related samples
Quantitative Reflectance Spectroscopy � Useful for mining process optimization Real-time or near real-time knowledge of mineralogical and metallurgical properties that impact metal recovery, allows for ▪ Intelligent ore sorting ▪ Optimization of ore processing � Useful for gangue minerology to minimize process cost and increase yield Gangue can affect extractability ▪ Talc and hornblende interfere with flotation ▪ Carbonates increase acid costs ▪ Clays can reduce yield due to loss of heap permeability
ore. Xpress Mineral Analysis/ Identification www. spectralevolution. com
Iron Minerals www. spectralevolution. com
Calcite Talc Hornblende www. spectralevolution. com
Clays www. spectralevolution. com
Reflectance Spectroscopy Advantages of reflectance spectroscopy � High throughput Hundreds to thousands of samples per day – ideal for rapid blast hole chip analysis Frequent (<1 minute intervals) measurements for in-process sensors � Non-contact measurements � Simultaneously determine multiple properties
Calibration Process Create Standards Collect Spectra Build, Optimize & Test Model Access Model Measure Unknown Predict Concentrations
Step 1: Prepare Set Prepare Calibration Set � Samples should reflect the physical properties and diversity that will be encountered in the field � Analyze the properties of interest using appropriate reference analytical methods, such as: Qemscan X-ray diffraction Acid consumption Other metallurgical tests � Measure the reflectance spectra
Step 1: Prepare Set Things to consider in measuring spectra � Features can overlap and may not be from a single component � Spectral features in minerals can result from crystal field effects, charge transfer, color centers, and conduction band transitions � Spectral features in organic and industrial samples come primarily from CH, NH, OH, and SH bonds � Multivariate models can consider all, or a substantial portion of the whole spectrum
Step 2: Develop & Validate Develop and validate your calibration � Match each reflectance spectrum you have collected to the corresponding reference analyses � Develop calibration equations using multivariate chemometric techniques like partial least-squares regression � Validate the performance of the calibration by using an independent set of samples
Step 3: Reference Method How to select a reference method � NIR is a secondary method – the reference needs to be well controlled with the lowest possible error � The Standard Error of Laboratory (SEL) should be known and documented � If there are changes in the reference method, new reference data may be substantially different from your original data � Submission of known samples is a good idea
Step 4: Spectral Collection Things to consider in collecting spectra � Verify your system performance using wavelength standards � Control particle size, moisture, temperature, and sample packing , or stabilize your model to resist changes in these parameters � Use the sample preparation as optical geometry can affect your outcome
Step 5: Apply Calibration Now apply your calibration � Prepare unknown samples with the same method used for calibration samples � Measure the reflectance spectrum of the unknown using the same set-up used in building the calibration � Apply the calibration to the unknown reflectance spectrum to predict mineralogical and metallurgical properties
Number of Samples How many samples will I need for calibration and test? � Reserve 20% of samples for an independent � 60 -90 samples for a feasibility study � 120 -180 for starting model � >180 for a robust production model test set
Calibration & Validation Samples How many samples will I need for calibration and validation? � Cover the anticipated range of composition � Scan in the form that will be analyzed by the model – make them match � Contain a natural combination of minerals - avoid blending as it can cause problems, beware cross correlations
Maintaining the Model Ensuring that your model retains its integrity � Watch out for samples with high spectral residual and samples that predict at or near the extremes of your model � Establish a consistent monitoring program with reference analysis done frequently � Implement a plan and schedule for improvement of the model including identifying new samples � Establish criteria for revising the model based on time, increased validation error, or similar characteristics
Examples of chemometric analyses using reflectance spectroscopy
- Journal of chemometrics
- Daniel spielman spectral graph theory
- A brief introduction to spectral graph theory
- Difference between strip mining and open pit mining
- Web text mining
- Strip mining vs open pit mining
- Strip mining before and after
- Mining multimedia databases in data mining
- Mining complex types of data in data mining
- Spatial data mining applications
- Data mining in crm
- Factors affecting width and intensity of spectral lines
- Spectral regrowth
- Spectral regrowth
- Spectral classification
- Profil spectral rigel
- Spectral normalization gan
- Spectral hashing
- Spectral efficiency
- Domaine spectral
- Spectral leakage
- Spectral bands
- Eric xing
- Spectral clustering
- Vsc 80
- Spectral clustering
- Spectral characteristics of angle modulated signals
- Analytical spectral devices