In this joint webinar, Jason Knuth, data scientist and analytics lead at Komatsu shares how they are analyzing over 17 billion data points every day from connected devices and using machine learning and analytics to improve mining operations.
2. Your Speakers Today…
Vijay Raja
Solutions Marketing Lead, IoT
Jason Knuth
Product Manager, Analytics and Smart Solutions
3. IoT Data Characteristics
- Why this is a Big Data andAnalytics Problem?
IoT data comes from a variety of different sources
• Massive volumes of intermittent data streams
• Generated from a variety of data sources
• Predominantly time-series
• Can come in streams (real-time) or batches
• Diverse data structures and schemas
• Some of it may be perishable
Combining sensor data with contextual data is the key
to value creation from IoT
5. Cloudera Enterprise – The Data Mgmt. Platform for IoT
Connected
Devices/ IoT Data
Sources
Internal Systems External Sources
BI Solutions Real-Time
Apps
Search Data Science
Workbench
SQL
Machine
Learning
Data Center
Hybrid
Cloud
Sensor/ IoT Data
• Data Storage
• Data Processing
• Machine Learning
• Real-time Analytics
OPERATIONS
Cloudera Manager
Cloudera Director
DATA
MANAGEMENT
Cloudera Navigator
Encrypt and KeyTrustee
Optimizer
BATCH
Sqoop
REAL-TIME
Kafka, Flume
PROCESS, ANALYZE, SERVE
UNIFIED SERVICES
RESOURCE MANAGEMENT
YARN
SECURITY
Sentry, RecordService
FILESYSTEM
HDFS
RELATIONAL
Kudu
NoSQL
HBase
STORE
INTEGRATE
BATCH
Spark, Hive, Pig
MapReduce
STREAM
Spark
SQL
Impala
SEARCH
Solr
SDK
Partners
Other Enterprise
Data Sources
6. Powering a Variety of IoT Use Cases…
Connected Vehicles
Usage Based Insurance
Industrial IoT Smart Cities & Ports Energy & Utilities
Smart HealthcarePredictive Maintenance Aerospace & Aviation
7. 7
Jason Knuth
Product Manager - Analytics & Smart Solutions
Komatsu
@Dataguy21
The Digital Mine
How data and analytics are
optimizing mineral extraction
9. 9
Company overview
Smart Solution Centers
• Around the clock monitoring
• Custom Data Solutions
• Productivity analysis
• Training
• And more…
10. 10
Mining market challenges
Mines are driven by cost per ton
Today’s challenges:
• Increasing social and regulatory issues
• Difficult mining conditions – deeper ore
deposits
• Declining commodity prices
• Inventory management
• Aging workforce
11. 11
New challenges need new solutions
Move from equipment-centric to holistic solutions
- Smart equipment with built-in sensors
- Advanced services beyond field repairs
- Data monitoring to manage machine health
- Optimize mine performance
- Lowest cost per ton
Monitoring
Continuous Improvement(CI)
Repair &Maintenance (R&M)
Productivity
(ton/h)
Production
Cost
($/ton)
TCO
($/h)
12. 12
Smart Solutions Definition
Smart Solutions are
integrations of our smart
connected products and
systems, advanced analytics
and direct services
customized to solve
customers’ toughest
challenges.
• Current offering of smart connected products: Shovels,
Wheel loaders, LHD, Drills, Draglines, Trucks, Longwall systems
(shearer, AFC, PRS, Pump stations) Conveyors, Continuous Miners,
Roof Bolters, Feeder breakers, FCT, Shuttle car, Battery Hauler
• Technology products: Controls and drives, environmental,
Mining intelligence, Operator assist, Proximity detection• 24/7 support in
over 20 countries
• Application engineers
experts
• Machine assembly
and rebuilds
• Life Cycle Management
• Component Exchange Program
• Customized training
• Inventory management
• Service products
and consumables
• Technical and field services
• Prognostics and
health monitoring
• Sub-second data
• Komatsu analytics
• Operational Excellence
• Service Excellence
Smart Solutions
14. 14
Our IoT Analytics Journey
Future Opportunities
• Cloudera PaaS - Altus
• End-to-end data
governance
• End-to-end data security
• Edge analytics
Gen 1 Gen 2 Gen 3 Gen 4
• Limited analytical
capabilities
• Data processing and
storage costs
• Limited machine
learning capabilities
• Data silos
Increasing value and business impacts
Historian Based
Out-of-the-Box
Proprietary Solution
Cloud based Open
Architecture & Platform
Driving Intelligence
at the Edge + PaaS
• Limited flexibility/
interoperability
• Scaling to data growth
was a challenge
• Data processing and
storage costs
• Challenges around
custom use case
Current State Future State
• Cloudera – Cloud
ecosystem
• Processing, analytics
and machine learning
for IoT data
• Processing 200,000
data points/second
• Rapid prototyping
3-2 Years Ago9-3 Years Ago
15. 15
Managing the data
This machine:
1,250 sensor points
550 alarms & events
All machines:
17 billion time series data
points with 3xPeak
6 million Alarms & events
150k analytics executed
Each day
18. 18
Data value chain
Decision Making
• Reporting & dashboarding
• On-time decision making
System Optimization
• Real-time asset monitoring
• Service lifecycle management
• Lowest cost per ton
Continuous
Improvement
• Support optimization
• Product use enhancement
Sales & Marketing
• Market intelligence
• Sales forecasting
Improved Logistics
• Supply chain optimization
• Production forecasting
Product Development
• New product development
• Product enhancements
• Product resolution
Mine learnings ● Essential partnerships
● Solutions aligned to customer goals ● Systems approach
Enhanced equipment● Customized solutions
● Optimal machine availability ● Integrated parts and service support
19. 19
Analytics Example--Problem Identification
Mineral extraction is rapidly becoming more
challenging - increasing the duty cycle of the
traction gearbox
Our onsite service teams noticed an increase in
gearbox anomalies
Dynamic Steady
Temperature
20. 20
Analytics Example-- Modeling and Evaluation
As OEM, leverage engineering-based modeling to create new feature
Each machine is unique signature, digital fingerprint
Engineering has since redesigned the gearbox for the new conditions closing the
data value chain.
The 3rd generation platform enables analytic development from months days
Irregular Normal
3-5 Days
7 Days
1 Day
21.
22. 22
The digital mine
Focused on optimizing mineral
extraction for the least impact
- Safety
- Autonomy, increased reliability
- Environment
- Less waste, improved efficiency
- Minimal footprint
- Productivity
- Continuous improvement
- Bringing lean practices to mining
- Maintenance
- Less wear and tear on the machines
- Reducing unplanned downtime
- Reducing maintenance time
GM – welcome to the session on IOT
Vijay Raja – Solutions Marketing Lead for IoT
Dave Shuman –
In the next 50 minutes we’ll try and dive into
How Cloudera fits in the IoT Ecosystem
What are some of the key enabling capabilities specifically around IOT
How we are working with partners to drive an end-to-end architectures for IOT
More importantly, we will talk about use cases
We’ll talk about what Customers are doing with Cloudera on IoT today
We will walk through a number of customer case studies and examples
Hopefully we will have some time towards the end for some Q&A as well…
IoT will generate the volume and variety of data greatly exceeding those with which information leaders are familiar with today, requiring modernization of information infrastructure to realize value.
Some of the key characteristics of IoT data include:
Massive volumes as we saw in the previous slide
But it is not just about the volumes that organizations need to be prepared for
We are talking about a wide variety of data sources, structures, schemas and formats – from sensor readings of temperature and pressure at one end to doing video analytics on live video streams
Most importantly it may come in streams (real-time) for lot of the IOT use cases or in batches – You need a data management platform that can manage both data-in motion and data at rest
Most importantly, the real value from IoT comes when you can combine that sensor data with other internal and external data sources – adding context to that sensor data
Let’s take a look at how Cloudera’s EDH fits into the IoT Ecosystem
Can ingest data from multiple sources including real-time streaming sensor data
You can combine the sensor data with data other internal and external sources to drive business insights
You can deploy EDH on prem (in your data center) or on hybrid cloud environments and still be able to manage it centrally
And you can serve and analyze the data in a number of different ways - integrate it with existing BI solutions, do SQL queries with impala, or do machine learning or integrate it with real time applications (such as OnCommnad Connect in the case of Navistar)
Our equipment and service offerings cover a full range of mining needs and can be combined to provide full-system solutions for surface and underground mining. Integrated services and solutions help customers
Serve customers for the life of the machine
MarComm
Read the definition
Dissect the graphic:
“We realize the definition is fairly long and cumbersome, so the intention of this graphic is to simplify the understanding of what JoySmart Solutions are. At Joy Global, we have been offering smart products, direct services and analytic services for a while independently. This new approach, called JoySmart Solutions has the intention of integrating these 3 main offerings with the objective of delivering value to the customer. The key aspect here that differs from everything else we have been doing until today, is that we are partnering with our customers to understand their challenges, and will offer them an integrated solution of smart products, direct service and analytics that will drive to an specific result, which will move the needle. Hence the colored dashboard on the center of this graphic. JoySmart Solutions are the means to an end result.”
Onsite team contacted our Smart Solution Analytics teams to pull data prior to the anomaly to determine if they could detect the event prior to need for manual intervention
Through data preparation multiple machine sensors were identified as possible predictors, additionally from our domain knowledge of our equipment we created additional features based on engineering principles.
We Applied the Principle Component Analysis (PCA) to the 6 traction gearbox Sensor data.
We utilized a random sample of 50% of the data from the first 40days assuming this will represent “normal” operation
Calculated the principle components and their coefficients to be used to score the full data set
After further evaluations of the axis coefficient’s it was determined that 2 of the engineered features were the best at describing the irregularity in the gearbox. In the end we utilize a Gaussian Mixture Model of the 2 features to describe the machine's digital finger print. Since conditions can vary from customer to customer we apply a digital finger print to each machines. Live data coming from the machine is evaluated against its own model.
Our engineering team evaluated these models and conditions to determine the best methods to redesign the gearbox for the customers tougher conditions. Since then these gearboxes have been deploy to tackle these new conditions. This is a good example of how we close the data value chain that I spoke about before this example.
Our 3rd generation platform enables us to develop analytics and models like this example in days rather than months when comparing to our previous platforms. The key driver in the improvement was breaking down the data silos and enabling quicker data wrangling and feature creation.
We are excited to see the rapid development of the IoT space with new technologies to enable us to deliver these types of analysis to our customers. Here is a video highlighting the path forward for Komatsu.
There’s no out-of-the-box solution for mining, but every solution we create in today’s digital mining environment focuses on …
Data is critical part of our journey to a more sustainable, less impactful mining environment
Still allow us to produce the essential minerals needed for modern society in a more efficient way