Particle Size Measurement of Crushed Ore Using 3D-profile Measurement
The particle size distribution of the crushed ore plays a significant role in autogenous grinding process. In Pyhäsalmi, the estimation and prediction of the incoming particle size distribution is valuable information for the mine. The measurement is carried out by combining mass flow- and 3D-profile-measurements.
Pyhäsalmi Mine Oy, owned by the Canadian Inmet Mining Corporation, operates a copper, zinc and pyrite mine in central Finland. The Pyhäsalmi mine is the oldest operative mine in Finland and one of the deepest in Europe since the new shaft reaching 1430 meters at depth was opened in the summer of 2001. The mine production has steadily increased and is currently about 1.3 million tons per year. The Pyhäsalmi mine is considered to be one of the most cost-effective and best performing underground mine world-wide. The blasted ore is first crushed in the bottom of the mine in order to achieve fluent and reliable operation of the ore transportation system. This first crushing stage defines the particle size distribution that is fed to the grinding mills in the surface. Since the transportation delay is typically two to three days, a measurement system that is capable of measuring the size distribution already in the mine was developed. Measurement is carried out right after the jaw crusher and the location of the measurement point is indicated in the Figure 1 below.
Because the size distribution is readily available in the mine it can be used to predict changes in the surface up to three days in advance. This helps the mine personnel in production planning and ensures that the amount of different size classes in the surface are kept in allowed range. Eventually this leads to economical savings in iron ball consumption, which have to be used in replacement of missing grinding media i.e. rocks of correct size. The mine personnel has estimated that out of 1 000 000 EUR used annually for the iron balls, some 20% could be saved if there would be correct size classes in the silos at all times. Measurement System The measurement system (see Figure 2) consists of the following components:
The belt weigher is equipped with a speed sensor and is thus capable of measuring the mass flow under the scanner. This information along with the speed reading is transferred as a standard 4-20mA message to the A/D-board of the analysing computer. Since the laser scanner is located on a same cross-directional axis as the belt weigher, the data coming from these two devices can be easily combined. The laser scanner is connected to the extra NIC-card of the computer via standard Ethernet connection. Calculated results are stored in a local database that can be accessed by using a client-software written for this purpose. In addition, the results are automatically uploaded to the automation system, enabling the plant operators a direct access to the results.
Analysis of measurement data
Estimation of particle size distribution from the 3Dprofile provided by the laser scanner is done in three steps. First, visible ore particles are extracted from the 3D-profile (Figure 3). After that particle size distribution is calculated for these recognized particles. In the third step this distribution is used for estimating the particle size distribution of the whole ore mass.
Ore particles are recognized using a combination of commonly known image segmentation techniques such as watershed segmentation and different morphological operations. Two different seed images are generated for the watershed segmentations which carry out the separation of particles from the background and separation of individual particles from each other. The resulting two segmentations are then combined in order to acquire the final result.
Since only a small portion of the whole ore mass is visible on the surface (10 to 30 percent of total volume) it is necessary to somehow estimate the distribution of the hidden portion of the mass. This distribution is currently estimated using the distribution of the visible part, but also other features such as texture properties of the surface could be used. A linear regression model with additional nonlinear input variables is used for the calibration. The nonlinear input variables are needed in the estimation of the submerged rocks and they are formed separately for each size class.
The particle size measurement system has been calibrated and tested using 32 samples, each sample being 1.5 meter long block of ore from the conveyor belt. These samples have been screened manually to the size classes used on the surface for controlling the grinding process. These classes are; fine ore (0-35mm), pebbles (35-100mm) and lumps (>100mm). Since the number of samples is barely adequate for the determination of the calibration model parameters, the analysis system have been tested using leave-one-out cross-validation method. The correlation between estimated and real values is demonstrated in Figure 4. The average absolute values for the errors are 4.55 kg for fine ore, 4.33 kg for pebbles and 4.34 kg for lumps. Average total mass of the samples is 143 kg. This accuracy level is good enough for this application. Figure 3. Flow chart of the particle recognition algorithm
These first results look very promising and a long term data collection campaign will be performed in order to validate the system more thoroughly. This can be carried out without the hard and time consuming manual labour by comparing the final results from the sieving station on the surface to the ones acquired already in the mine. However, the cumulative amount of ore fed through the system must be high enough to diminish the disturbances from the transportation system.