SLIDE 1
Mineral processing
2
Machine learning in mineral processing Lidia Auret Process - - PowerPoint PPT Presentation
Machine learning in mineral processing Lidia Auret Process - - PowerPoint PPT Presentation
Machine learning in mineral processing Lidia Auret Process Engineering 23 March 2018 Maties Machine Learning Mineral processing Continuous, connected, controlled, circulating, complex, changing 2 Industrial data Online physical
SLIDE 2
SLIDE 3
Industrial data
- Online physical property sensor data
– E.g. mass flow rate, density, temperature, pressure – ~ seconds
- Online image data
– E.g. rocks on conveyor belts, flotation froth (mud and bubbles) – ~ minutes
- Offline laboratory data
– E.g. metal content, particle size distribution – ~ hours
- Offline image data
– E.g. microscopic grain shape and colour – ~ days
3
SLIDE 4
Abnormal event detection
4 Disturbance processes D
- Many faults and failures can occur in complex processes
- Large variation in normal operating conditions due to
range of allowable disturbances
SLIDE 5
Abnormal event detection
5
- Missed detections can lead to suboptimal performance,
equipment failure, safety and environmental violations
- False alarms can lead to unnecessary downtime and loss
- f trust in alarm systems
SLIDE 6
Abnormal event detection
6
- Unsupervised learning problem
– Abundance of one class of data: Normal operating conditions
- Fault detection
– Feature extraction – Data description / support estimation
- Fault identification
– Topology extraction – Supervised learning model inspection:
- Variable importance
- Partial dependence
SLIDE 7
Abnormal event detection
7
- Feature extraction
– Sensor data correlated (through mass and energy balances, control instructions) – Sensor data noisy – Feature space represents lower dimensional, noise-free information – Residual space represents feature extraction model validity
SLIDE 8
Abnormal event detection
8
- Feature extraction
– Principal component analysis
- 𝑼∗ = 𝒀𝑸∗;
𝒀 = 𝑼∗ 𝑸∗ 𝑼
– Kernel principal component analysis
SLIDE 9
Abnormal event detection
9
- Feature extraction
– Autoassociative neural networks (NLPCA)
SLIDE 10
Abnormal event detection
10
- Feature extraction
– Autoassociative neural networks (NLPCA)
SLIDE 11
Abnormal event detection
11
- Data description / support estimation
– Kernel density estimation – One-class support vector machines
SLIDE 12
Abnormal event detection
12
- Data description / support estimation
– Kernel density estimation – One-class support vector machines
SLIDE 13
Abnormal event detection
13
- Data description / support estimation
– Kernel density estimation – One-class support vector machines
SLIDE 14
Abnormal event detection
14
- Topology extraction
– Identification of propagation path of fault – Transfer entropy / lagged cross-correlation used to determine direction and strength of connections between variables
SLIDE 15
Abnormal event detection
15
- Topology extraction
– Identification of propagation path of fault – Transfer entropy / lagged cross-correlation used to determine direction and strength of connections between variables
SLIDE 16
Abnormal event detection
16
- Topology extraction
– Identification of propagation path of fault – Transfer entropy / lagged cross-correlation used to determine direction and strength of connections between variables
SLIDE 17
Abnormal event detection
17
- Research approach
– Scarcity of industrial data with faults detected and identified – Simulation of complex, dynamic processes with known faults – Repository with dynamic models and simulated data
github.com/ProcessMonitoringStellenboschUniversity
SLIDE 18
- Ore characteristics
– Metal content, particle size correlated to process performance – Captured by image data
- Soft sensor
– Trained model for prediction of process performance from measured process data
Soft sensors
18
GLCM Wavelet Textons Steerable pyramids Etc. k-NN LDA Ore grade Particle size Process state
SLIDE 19
- Flotation grade prediction with convolutional neural
networks texture features and classification
Soft sensors
19
SLIDE 20
- Flotation grade prediction with convolutional neural
networks texture features and classification
Soft sensors
20
Variants generated to supplement data set Class labels used for training Bottleneck introduced to create lower dimensional feature space
SLIDE 21
- Flotation grade prediction with convolutional neural
networks texture features and classification
Soft sensors
21
SLIDE 22
- Flotation grade prediction with convolutional neural
networks texture features and classification
Soft sensors
22
Interpretability important for industrial adoption
SLIDE 23
- Data size, quality and fusion
– Potentially massive data sets – Shifting process conditions – Online process data + offline process data + maintenance records + mine plan + purchase orders + etc.
- Exploiting process knowledge
– Dynamic Bayesian networks – Hybrid modelling
- Process recovery
– Actionable insights – Reinforcement learning
Challenges and opportunities
23
SLIDE 24
Questions?
24