BigML’s partners, A1 Digital, introduce how the Internet of Things and Machine Learning can bring business value in Logistics.
Speaker: Francis Cepero, Head of Vertical Market Solutions at A1 Digital.
*Intelligent Mobility 2021: Virtual Conference.
Biometric Authentication: The Evolution, Applications, Benefits and Challenge...
Intelligent Mobility: Business Value of IoT and ML in Logistics
1. Business Value of IoT and
Machine Learning in Logistics
Francis Cepero
Director Vertical Markets
2. 2
B2B Partner in
Cloud, IoT, ML and Security
+1500 Customers
+500 international IoT
customers
Gartner MQ IoT Services 2020
Corporate Group
100% Subsidiary of A1
Telekom Austria Group,
Part of America Móvil
200 Employees
3 locations in
Central Europe:
Vienna, Munich
and Lausanne
A1 Digital in a nutshell
Intelligent Mobility 2021 - BigML - A1D
3. 3
6 Classic Problems in Logistics
Asset profitability impacted
from inefficient planning and
operation (idle times).
Asset availability impacted by
maintenance and down time
Lack of visibility on the daily
operations, few data enabled
metrics or key performance
indicators
SERVICE OPERATIONS
Lack of optimized usage of
assets, integrated planning,
route planning
No scale due to lack of
operational capacities.
Service level compliance
issues and associated
penalties.
Inability to adapt to
customers growing demands
CUSTOMER
ENGAGEMENT
Few insights through data
analysis (e.g. machine
learning models for demand
and supply chain)
Strong dependency on
manual repetitive tasks.
Lack of automated and
optimized planning capacities
to attend growing demand.
MANUAL PLANNING
ASSET OPERATION OPTIMIZATION DEMAND VISIBILITY
Intelligent Mobility 2021 - BigML - A1D
4. 4
Scenarios Quality Control Maintenance Operations
Descriptive -
what happened?
• Quality Monitoring
• Testing Process
• Monitoring & Evaluation
• Detect Quality Loss
• Equipment Monitoring
• Performance Analytics
• Maintenance Analytics
• Equipment Failure RCA
• Operations Monitoring
• Process Mining
• Operator Behavior
• Operation Failure RCA
Predictive -
what will happen?
• Early Defect Detection
• Yield Quality Predict
• Predict Failures
• Estimate remaining useful
life (RUL)
• Predict Failure Impact
• Predict Activity / Setup Times
• Predict Production KPI(s)
• Demand Forecasting
• Supply Chain Disruption
Prescriptive -
what to do?
• Process Parameter
Recommendation for
Quality Improvement
• Self-calibrated testing
• Reduce Failure Cost
• Reduce Failure Rate
• Repair Recommendation
• Optimize Maintenance
• Start parameters
optimization
• Failure Rate Reduction
• Fuel/Energy Reduction
• Equipment Scheduling and
Dynamic Dispatch
• Operations Recommendation
Opportunities for compound improvement with IoT/ML
Smart Assets Use Cases (sense/predict/react)
RCA: root cause analysis
5. 5
Predictive Analytics for Rail Logistics
Predictive Analytics future – integrating for success
Level 0: Prepare for business impact
Select Pilot Use cases, Collect data, select partners, align on platforms, run first PoCs, first
Business Cases, test highest value with minimal risk.
Level 1: Subsystems: Predictive analytics is executed at the subsystem /edge level
Use Cases: wheel bearing damage, flat spots, weight detection, exhaust filter, pantograph
Level 2: Asset Management: Condition of the cars / devices
Use Cases: Detect anomalies and combine them with asset data.
Condition/Predictive Maintenance of single asset types
Level 3: Operations Management
Use Cases: Multimodal logistic, predictive demand, supply chain simulation, predictive repair
cycles, scheduling, optimizing shifts, call centers, secondary assets
Level 4: Strategic Management
Use Cases: Strategic Investment decisions based on capacity utilization, market analysis,
economic activity (mega/macro/micro cycles), strategic decision support systems
Platform based approach: sustainable and repeatable impact at low costs.
Avoid lock-in. Achieve economies of scale, scope and skills at business, technical
and commercial levels
Enterprise
Systems
ERP
EAM
SCM
CRM
LPS
external
…
DWH
BPM
Strategic
Initiatives
Innovation
Programs
Continuous
Improv.
PDCA
Balanced
Scorecards
Quality
Control
Programs
Operations
Mgmt
Platforms of differentiation Systems of records
ML
Models
ML
Datasets
Clusters
Assoc.
sources
Anomaly
IOT
Device
Management
Real-Time
Analytics
Data
Visualisation
Integration
Device
Connectivity
Storage
Platform based predictive applications
6. 6
Real Time Digital Logistics: Connected Assets and Connected Planning
Intelligent Mobility 2021 - BigML - A1D
IoT Solutions
+
Advanced Logistic Planning
Source: A1Digital/MathITLogistics
real-time management on assets and shipments real-time modelling of transportation networks
8. 8
Target: Manage all type of assets and tools centrally - at a much lower cost
Locomotives/Heavy
Machines/Passenger Trains
Rolling Stock/Freight Cars
Other Vehicles/Forklifts/
Containers/…
Machines
Rail Tracks/Switches/
Railroad Crossings
Tools,
Workshop Equipment,
Spare Parts,
…
Facilities
Shunting/Logistics/
Disposition
Humans, Experts, Teams
Infrastructure/CheckPoints/
Visual Detection/Damage Recognition
Intelligent Mobility 2021 - BigML - A1D
9. 9
Integrated Asset Management and IoT
Do this: EAM for Integrated Maintenance and Asset Optimization
• Modern enterprise asset
management (EAM) system to
improve processes and customer
service
• Identify opportunities to reduce
asset maintenance expenses with
better visibility into costs
• Manage planned and corrective fleet
maintenance more efficiently
• Fast transition (six-month)
• Fast integration with IOT based
solutions
EDGE ML PROCESS ML
13. 13
Can we understand complex data streams? Can we predict them?
ML predicts the fR_Mean with an excellent accuracy
14. 14
ML Workflow
1. ML task: predict fR_Mean (a number) → Regression (supervised learning) with metric R2 to quantify prediction accuracy
2. Variety of ML models for Regression e.g Decision Tree (DT), Random Forest (RF), Neural Networks etc
3. Each model has several tuning parameters e.g #nodes for a DT, #Trees for RF
4. We examined systematically (AutoML)~ 30 models with all possible parameter combinations and compare R2
15. 15
Integrate IoT and ML with modern Asset Management Systems to add value. Many
important scenarios beyond predictive maintenance …
Key take-aways for Connected Logistics
1
2
3
IoT enabled Asset Management and Logistics Planning will create additional business
value in your organization and improve your customer acquisition costs.
IoT = Team sport: Partnership between business and tech experts always helps!
Intelligent Mobility 2021 - BigML - A1D