XERVON Innovation Case

How XERVON Functionalizes Machine
Learning for Predictive Maintenance
12/16/2019©2018itelligence

The Case: Predictive Maintenance Based on IoT and ML
12/16/2019
2
Challenge
Running cooling towers account for 70% of the electricity costs.
The challenge is to operate them by adjusting pumps and fans in
an optima way, save energy and maintenance costs based on
contractual conditions. A process based on best class technology
is to be implemented which enables predictive operation and
optimizes for the lowest possible infernal power consumption.
©2019itelligence
Solution
An end-to-end scenario based on delivery of sensor data, data transfer, analysis and monitoring with an
integrated machine learning model which enables predictive operation and optimizes for the lowest possible
internal power consumption. The contractual conditions (cooling capacity, flow and line pressure in winter and
summer operation) must be observed.
Outcome
Digital twin taking care of optimized cooling water supply, energy savings, optimized pump and fan operations,
predicts maintenance and the optimal operarating state for each accepted cooling demand.

Facility Management Based on Sensors
12/16/2019
3
©2019itelligence
 Control of pump variables like cooling capacity, flow and
line pressure in winter and summer operation
 Pump characteristics have created a nonlinear course of
added complexity
 Modernization of the cooling tower with automation
technology
 Algorithm intervenes in control to set each optimum
operating state for each accepted cooling power
 Automatic condition monitoring of rotating machine parts
based on existing data (histories) and alerting in case of
deviations of the trained behaviour.
 automatic early fault detection mechanism to prevent
primary damage and consequential damage and thus
downtime.
 Spectra, storage values and characteristics are part of the
extensive analysis
Cooling Towers Condition Monitoring

Cooling Towers
12/16/2019
4
©2019itelligence
 Control of pumps consisting of 3
existing pumps with 6kV motors
(without FU) and 2 pumps with
FU (minimum speed must be
50%)
 Control variables are cooling
capacity, flow and line pressure
in winter and summer operation
 Pump characteristics have
created a nonlinear course of
added complexity
 Modernization of the cooling
tower with automation
technology
 Algorithm intervenes in control
to set each optimum operating
state for each accepted cooling
power
 Data Basis > SQL Server
 Scalability > other locations
 Automatization > data transfer
 Correlations > known
 Energy saving delivers cost
savings
 Wear reduction due to less
frequent start-up
 Intelligent distribution of
operating hours on the existing
pumps
 Integration of weather data
 Flow measurement available
(redemption after acceptance)
 By only standard DIN EN ISO
50001
Project Target General ConditionsBenefits

Condition Monitoring
12/16/2019
5
©2019itelligence
 Automatic condition monitoring
of rotating machine parts based
on existing data (histories) and
alerting in case of deviations of
the trained behaviour.
 The aim is an automatic early
fault detection mechanism to
prevent primary damage and
consequential damage and thus
downtime.
 Spectra, storage values and
characteristics are part of the
extensive analysis
 The evaluation in the field of
condition monitoring should be
speeded up and automated. In
the field of online systems, the
algorithm serves as an alarm
system.
 Data Basis > Encrypted SQL
database
 Scalability > unlimited
 Automatization > online and offline
measurements
available
 Correlations > known
 Automated evaluation of large
amounts of data
 Potential for connection to
online measuring systems
 New service for the? Condition
monitoring market
 Correlation of vibration data
with other parameters possible
Project Target General ConditionsBenefits

A 360 Degree View of a Digital Twin
12/16/2019
6
©2019itelligence
A 360 degree view across asset model, master,
transactional and performance data
Information
 Highlight Cards, Data Sheets, Equipment/Model Information,
Installation Location, Business Partners
Structure and Parts
 Structure, Spare Parts, Visual Parts
Documentation
 Documents, Instructions, Failure Modes, Alert Types,
Announcements, Improvement Requests, Fingerprint
Monitoring
 Alerts, Indicators, Component Indicators, 2D Charts
Maintenance and Service (integration)
 Notifications, Work Orders, Tickets, Contracts
Analytics
 Failure Mode Analytics
Model/Equipment History/Timeline

Advanced Analytics to Support Maintenancen Execution and Strategy Decisions
12/16/2019
7
©2019itelligence
Failure Mode Analytics
 Utilizes machine learning to generate KPIs around
documented failure modes
Fingerprint Management
 A visual approach to capturing asset reference
states. Used to visual comparison to current
operating performance (i.e., trend analysis)
IT/OT Data Fusion Views
 Equipment lists and geospatial views combining
model data and sensor based health indictors to
prioritize maintenance actions and support
strategy decisions

Intuitive and Scalable Machine Learning
12/16/2019
8
©2019itelligence
Dataset Configurator
 Intuitive dataset preparation capabilities to
prepare data for machine learning (i.e.,
aggregation periods, new features, null values)
Automated Machine Learning
 Run automated anomaly detection & failure
prediction algorithms without the need to choose
algorithm or hyper parameter settings
Flexible Extension Concept
 Ability to deploy custom algorithms into SAP
Predictive Maintenance and Service solution
Prediction & Anomaly Detection Algorithms
 Out-of-the-box machine learning algorithms
tailor made for PdMS use cases (i.e., developed
with industrial assets in mind)
Anomaly Detection Algorithms
Prediction Algorithms

IoT for Facility Management
12/16/2019
10
©2019itelligence
Systems
Monitoring
Messages
Analytics
 sensor data
 master data
Business data
 rules
 processes
 history
Applications
Machine data
Systems
SCP, Leonardo IoT
services, AE, PdMS, SAC,
ERP
Machine
Learning
Create an end-to-end scenario based on delivery of sensor data, data transfer, analysis and monitoring, an
integrated machine learning model to detect best operating parameters and triggering of follow up
processes like creation of maintenance order, send out alarm messages or maintain requests.
ActionMachines / Sensors

XERVON Innovation Case

More Related Content

What's hot (16)

Similar to XERVON Innovation Case (20)

Recently uploaded (20)

XERVON Innovation Case