In-process adjustment to crushing systems
Methods, systems, and apparatus, including computer programs encoded on computer storage media, for in-process adjustment to crushing systems. A method includes obtaining pre-crush particle data indicating characteristics of a portion of particles before crushing the portion of particles with a crushing system; obtaining settings data indicating one or more settings of the crushing system; obtaining post-crush particle data indicating characteristics of the portion of particles after crushing the portion of particles; and training a prediction model using the pre-crush particle data, the settings data, and the post-crush particle data, comprising: processing, using the prediction model, the pre-crush particle data and the settings data to obtain a corresponding output including predicted characteristics of the portion of particles after crushing the portion of particles with the crushing system; and adjusting parameters of the prediction model based on comparing the output of the prediction model to the post-crush particle data.
1 . A method comprising:
obtaining pre-crush particle data indicating characteristics of a portion of particles before crushing the portion of particles with a crushing system;
obtaining settings data indicating one or more settings of the crushing system;
determining, from the pre-crush particle data and the settings data and by a prediction model, predicted post-crush data that indicates predicted characteristics of the portion of particles at an output of the crushing system, wherein the prediction model is configured to receive both the pre-crush particle data and the settings data as input and predict how the crushing system will process the portion of particles based on both the pre-crush particle data and the settings data;
obtaining post-crush particle data indicating actual characteristics of the portion of particles at the output of the crushing system;
comparing the predicted post-crush data with the post-crush particle data to generate training data; and
training the prediction model with the training data.
2 . The method of claim 1 , wherein obtaining the pre-crush particle data comprises:
obtaining sensor data from a first sensor configured to capture low-resolution image of the portion of particles passing the first sensor at a high rate of throughput; and
generating the pre-crush particle data from the sensor data and by a characterization model configured to correlate low-resolution images with higher resolution images of particles and infer the pre-crush particle data based on the correlation.
3 . The method of claim 1 , wherein post-crush particle data is generated by a characterization model configured to determine particle characteristics from sensor data generated from measurements of the portion of particles after crushing the portion of particles with the crushing system.
4 . The method of claim 1 , comprising determining, using the trained prediction model, predicted post-crush characteristics of a second portion of particles by:
providing, to the prediction model, second pre-crush particle data indicating geometric and chemical characteristics of a second portion of particles before crushing the second portion of particles with the crushing system; and
receiving, as output from the prediction model, predicted characteristics of the second portion of particles after crushing the second portion of particles with the crushing system.
5 . The method of claim 1 , wherein the characteristics comprise geometric characteristics that include at least one of size, shape, surface area, or sphericity.
6 . The method of claim 1 , wherein the characteristics indicated by the pre-crush particle data include chemical characteristics that include at least one of chemical composition, mineral type, crystalline structure, or reactivity.
7 . The method of claim 1 , wherein the settings of the crushing system include at least one of material feed speed, working surface aperture size, or crusher operating speed.
8 . The method of claim 1 , wherein the prediction model comprises a neural network model.
9 . A method comprising:
obtaining pre-crush particle data indicating characteristics of a portion of particles before crushing the portion of particles with a crushing system;
obtaining settings data indicating one or more settings of the crushing system;
determining, from the pre-crush particle data and the settings data and by a prediction model, predicted post-crush data that indicates predicted post-crush characteristics of the portion of particles at an output of the crushing system, wherein the prediction model is configured to receive both the pre-crush particle data and the settings data as input and predict how the crushing system will process the portion of particles based on both the pre-crush particle data and the settings data;
determining an error between the predicted characteristics of the portion of particles and target characteristics for crushed particles output by the crushed system;
re-training the prediction model with training data generated based on the error; and
generating, from the error between the predicted post-crush characteristics and the target characteristics and by a control model, control signals for altering operations of the crushing system, wherein the control model is configured to receive as input the error between the predicted post-crush characteristics of the portion of particles and target characteristics.
10 . The method of claim 9 , wherein obtaining the pre-crush particle data comprises:
obtaining sensor data from a first sensor configured to capture low-resolution image of the portion of particles passing the first sensor at a high rate of throughput; and
generating the pre-crush particle data from the sensor data and by a characterization model configured to correlate low-resolution images with higher resolution images of particles and infer the pre-crush particle data based on the correlation.
11 . The method of claim 9 , wherein the settings of the crushing system include at least one of material feed speed, working surface aperture size, or operating speed.
12 . The method of claim 9 , comprising:
iteratively predicting post-crush characteristics of crushed particles output by the crushing system and adjusting settings of the crushing system until the error between the predicted post-crush characteristics of crushed particles and the target characteristics of crushed particles is equal to or less than a threshold error.
13 . The method of claim 9 , wherein the post-crush characteristics comprise geometric characteristics that include at least one of size, shape, surface area, or sphericity.
14 . The method of claim 1 , wherein comparing the predicted post-crush data with the post-crush particle data to generate training data comprises correlating pre-crush particle data with post-crush data associated with the same particles based on respective timestamps applied to the pre-crush particle data and the post-crush data.
15 . A system comprising:
at least one processor; and a data store coupled to the at least one processor having instructions stored thereon which, when executed by the at least one processor, causes the at least one processor to perform operations comprising:
obtaining pre-crush particle data indicating characteristics of a portion of particles before crushing the portion of particles with a crushing system;
obtaining settings data indicating one or more settings of the crushing system;
determining, from the pre-crush particle data and the settings data and by a prediction model, predicted post-crush data that indicates predicted characteristics of the portion of particles at an output of the crushing system, wherein the prediction model is configured to receive both the pre-crush particle data and the settings data as input and predict how the crushing system will process the portion of particles based on both the pre-crush particle data and the settings data;
obtaining post-crush particle data indicating actual characteristics of the portion of particles at the output of the crushing system;
comparing the predicted post-crush data with the post-crush particle data to generate training data; and
training the prediction model with the training data.
16 . The system of claim 15 , wherein obtaining the pre-crush particle data comprises:
obtaining sensor data from a first sensor configured to capture low-resolution image of the portion of particles passing the first sensor at a high rate of throughput; and
generating the pre-crush particle data from the sensor data and by a characterization model configured to correlate low-resolution images with higher resolution images of particles and infer the pre-crush particle data based on the correlation.
17 . The system of claim 15 , wherein post-crush particle data is generated by a characterization model configured to determine particle characteristics from sensor data generated from measurements of the portion of particles after crushing the portion of particles with the crushing system.
18 . The system of claim 15 , wherein the characteristics comprise geometric characteristics that include at least one of size, shape, surface area, or sphericity.
19 . The system of claim 15 , wherein the characteristics indicated by the pre-crush particle data include chemical characteristics that include at least one of chemical composition, mineral type, crystalline structure, or reactivity.
20 . The system of claim 15 , wherein the settings of the crushing system include at least one of material feed speed, working surface aperture size, or crusher operating speed.
21 . The system of claim 15 , wherein the prediction model comprises a neural network model.