Presented at AEMP Connect 2021 on March 10, 2021.
Can AI Solve Construction Telematics Overload Problem? Oded Ran, Co-Founder & CEO, Clue Insights (getclue.com) presents on the current problems construction companies face using telematics, is AI the solution, and what is the path forward.
AEMP Connect 2021 Can AI Solve Construction Telematics Overload Problem? Oded Ran, Co-Founder & CEO, Clue Insights (getclue.com)
1. Will AI Solve the Telematics
Overload Problem?
Oded Ran, Co-Founder & CEO
www.getclue.com
2. 1. What are the main problems using telematics today?
2. Is AI the solution?
3. What is the path forward?
Today we’ll talk about these three topics
3. 1. You’re involved in managing a heavy assets fleet
2. You’re over-worked
3. You’re wondering if there’s anything that can help you
or your organization
This presentation for you if:
9. But the main question is: is telematics data useful?
10. • Monitor asset health
• Send fault codes wirelessly
• Reduce downtime
• Improve performance
Telematics started with putting a mechanic in the cab
Credit: Prof. Mike Vorster
11. • Measure operating hours
• Idling hours
• Load counts
• Cycle time
• Improve utilization
…then adding a Production Engineer in the cab
Credit: Prof. Mike Vorster
13. • Telematics use algorithms to interpret outputs from sensors.
• They don’t actually know what it is the operational status.
Example: how telematics determines IDLING?
• Machine in gear and/or moving
• Engine RPM is high
• Implement is engaged
What actually are we measuring?
14. Is it idle or is it not idle?
Operational Status What Sensors Say
Not idle
Gear, movement, high RPM,
implement engaged
Idle
Not in gear, low speed or RPM,
implement not engaged.
Off
Engine off or no data
Credit: Prof. Mike Vorster
15. Is it idle or is it not idle?
Operational Status What Sensors Say
Not idle
Gear, movement, high RPM,
implement engaged
Idle
Not in gear, low speed or RPM,
implement not engaged.
Off
Engine off or no data
Working
Standby
Down
Credit: Prof. Mike Vorster
16. Is it idle or is it not idle?
Operational Status What Sensors Say
Not idle
Gear, movement, high RPM,
implement engaged
Idle
Not in gear, low speed or RPM,
implement not engaged.
Off
Engine off or no data
Working
Machine is likely working
and engaged in effective
work
Machine may be doing
contributory work (e.g.
excavator used as a brace)
or appears idle when it’s
not (e.g. paver at low RPM)
May happen due to
connectivity or fault
Standby
Unlikely to show as
‘non-idle’
Stand-by machines likely to
show as ‘idle’ if not
powered down
Stand-by machines are
likely to be powered
Down
No data except fault codes which may indicate a problem.
Credit: Prof. Mike Vorster
17. Which of these excavators presents a problem?
Credit: Duo UK (Clue Customer)
EX 1
EX 2
EX 3
EX 4
EX 5
EX 6
EX 7
EX 8
EX 9
18. Which of these excavators presents a problem?
Credit: Duo UK (Clue Customer)
a) EX1-EX3, as idle rate is > 50%
b) EX8, because fuel use is highest
c) EX9, because utilization is less than 25 hrs
d) All of the above
e) None of these
f) I have no idea
EX 1
EX 2
EX 3
EX 4
EX 5
EX 6
EX 7
EX 8
EX 9
19. Which of these excavators presents a problem?
Credit: Duo UK (Clue Customer)
a) EX1-EX3, as idle rate is > 50%
b) EX8, because Fuel Use is highest
c) EX9, because utilization is less than 25 hrs
d) All of the above
e) None of these
f) I have no idea
EX 1
EX 2
EX 3
EX 4
EX 5
EX 6
EX 7
EX 8
EX 9
20. Telematics is useless without context
80% of idling occurred
before shift start time
This is an 80-ton excavator; all
others are 30-40 ton
Note idling is lowest
Asset operated only
2 days as returned
from repair
Outside temperature
was -20F
New operator in training
Excavator used as brace
for 3 days
EX 1
EX 2
EX 3
EX 4
EX 5
EX 6
EX 7
EX 8
EX 9
21. Let’s hear it from Darrin
Darrin Sheriff
Director of Equipment & Purchasing
23. Artificial Intelligence (AI)
Intelligence demonstrated by machines
Machine Learning (ML)
Computer algorithms that improve through experience
The goal of AI & ML: make predictions
AI & ML – common definitions
28. AI is already showing promise in construction
Credit: Geotab
29. Anatomy of an AI project:
Getting live production data for 100ton haul fleet
30. Example: Live production data for 100ton haul fleet
Collect GPS data
every 6 seconds
(5-10k data points
per asset per day)
Getting the
raw data is
the easy bit.
31. Example: Live production data for 100ton haul fleet
Analyze speed, yards
moved, number of
minutes stationary,
plus encounters with
asset type ‘excavator’
or ‘loader’
32. Example: Live production data for 100ton haul fleet
Choose an AI
algorithm e.g.
K-means clustering;
DBSCAN, HDBSCAN
or a combination
Re-run model every
15 minutes to
provide real time
updates
33. Example: Live production data for 100ton haul fleet
Display load count in
applications;
send alerts when
below below target
Compare to manual
load count entries by
field in Oracle JDE and
daily field diaries
34. Example: Live production data for 100ton haul fleet
Display load count in
applications;
send alerts when
below below target
Compare to manual
load count entries by
field in Oracle JDE and
daily field diaries
36. 1. Digital data
2. Complete data
3. Connected data
4. Clean data
(And ideally, a good partner to help with this.)
Get your data ready for AI. You need to have:
37. • Inspections, WOs, Breakdowns and PM
• Production reports & daily diaries
• Timecards
• Materials, load tickets, truck tickets
• Operator/asset assignments
• Attachment used, unique machine specs
• Job site and project information, tasks and cost codes
• $ costs and rates
• Operator training & certification
• And any other data that can provide context
1) Digitize all data from field, machine & office
38. • Activate OEM TMS when possible
• Best & only reliable way to get
engine data
• Use after-market trackers (ideally
– one provider) to connect the rest
of your fleet
• Must provide useful operating data
• Minute-by-minute data is essential
2) Complete your telematics coverage to all assets
39. 3) Connect all systems & legacy data to a single
datastore - and move away from closed systems!
40. “Can you show me how to import any data into
your system and export all data out of it?”
3) Connect all systems & legacy data to a single
datastore - and move away from closed systems!
41. 4) Clean it, or find a partner that knows how
Credit: BuzzBoard
42. • Time (coordinating, entering, analyzing, reporting, & using data)
• Utilization (productivity, asset allocation, purchase decisions)
• Money (minimize down events, improper asset usage, optimize
replace/repair decisions)
• Employee satisfaction, brand value, competitiveness
It’s the “getting ready for AI” that will unleash huge
improvements for your organization
44. Thank you
Donte, Dave (AEMP)
Darrin and the team at Palmetto
Mike Vorster
Michael McLin (Maxim Consulting)
Justin Smith & Preston Ingalls (TBR)
Mike Branch (Geotab)
Clue team