Making Mining Digital
Digital vs. Analog
We are fundamentally changing mineral processing technology from an analog process using 19th Century chemical engineering technologies to a digital process based on the latest sensor and information technologies.
The basic idea behind our technology is that each particle in a powder can be measured, analyzed, and precisely positioned.
By grinding rock samples below the average size of the mineral crystals, most of the particles will be dominated by one mineral. At these small scales air pressure alone is strong enough to precisely position and direct the particles. Our micro-pneumatic channels are 200 microns wide – just slightly wider than the average thickness of a human hair of 150 microns. The small size of these channels allows us to work with the individual particles in a powder - a new capability with many applications.
How it Works
First we start with a 200 mesh powder (average particle size of 75 microns). This will involve crushing and grinding rock samples and feeding them into our device. The device works with a variety of particle sizes, but requires that particles are smaller than 200 microns in size.
Next, particles are put into single file and run across the sensor window. The sensor takes multiple measurements of the particle in order to compile an accurate spectral signature. Each mineral has a uniquely identifying spectral signature and this is why our device can work with any mineral.
Our sensor contains an onboard processor which can categorize each particle and trigger a switch which controls the air pressure in each micro-pneumatic channel. Once the right switch is triggered air pressure will drop in the selected channel and the particle will enter this channel and be deposited into a collection bin.
This process is repeated with each particle in the stream.
A full analysis of the sample is available with some post-processing of the input data and is usually available a few minutes after the sample run.
The device can be scaled up by adding additional particle channels on a custom designed chip with additional pixel lines.
The Portable Mineral Lab has only one input channel per sensor chip, but the Production Mineral Separator will have up to 500 input channels per chip.
Our first product is the Portable Mineral Lab which will be able to give the complete chemical composition of crushed rock samples in the field. Because each particle is measured we expect the sensitivities to be quite low and that minerals can be detected down to a few parts per billion. Current field technology can detect elements down to about 3 parts per million – usually elements only. Importantly, our device will be priced under $10,000 which is about 50% of the price of competing devices.
Our second big application is the Production Mineral Separator which will be allow operators to process almost any mineral concentrate. The initial module will process 1 ton per hour. Units are modular and can be added to increase throughput or increase the number of by-products. The Production Mineral Separator can serve as a finishing device or the primary separation method replacing current methods such as gravity or flotation.
In addition to mining applications our device will have many applications outside of mining and can detect organic molecules in addition to minerals with a software upgrade. Some of the non-mining applications include: paints and pigments, food contamination, powder coatings, adhesives, additives, metal powders, sediments and soils, cement, fertilizers, raw materials, household products, forensic chemistry, and law enforcement.
Intelligent Device Revolution
In his book, Cambrian Intelligence, the great robotics pioneer, Rodney Brooks, posited that intelligence was the intersection of perception and action as opposed to the operation of a master algorithm.
This type of intelligence – coordinated sensors and actuators – mechatronics – is really our core competence.
While better algorithms and data science will play an important role in the future of artificial intelligence – physical intelligence as embodied by sensor-intensive control systems will play an overlooked and possibly bigger role.
Novel combinations of sensors and actuators are easier to patent and more defensible than innovative algorithms. Furthermore, physical intelligence is massively under-invested compared to mainstream AI.