Download Hitachi PdM Solutions Enable Increased Asset Utilization

The emergence of the Internet of Things (IoT) as well as the need
for organizations to reduce costs, maximize asset productivity,
improve product or component quality and increase operational
performance, has led to organizations implementing predictive
maintenance solutions. The appeal is that these technologies,
which can be applied to various vertical markets, allow users to
determine the condition of in-service equipment and predict when
maintenance should be performed. Perhaps the most significant
purpose of Predictive Maintenance (PdM) is to perform that maintenance at a certain scheduled time when it can be most cost effective
and before equipment loses performance within a threshold.
The Hitachi Center for Social Innovation has developed a suite
of technologies to address Predictive Maintenance use cases for
solutions across verticals including, but not limited to: manufacturing, transportation and automotive, infrastructure management, oil
and gas, mining, and health. The data-driven technologies provide
insights that enable the maintenance staff to take the right actions
at the right times, which can help in avoiding unexpected failures,
decreasing maintenance costs and increasing equipment availability. Driven by advanced analytics, Hitachi PdM solutions can detect
anomalies and certain failure patterns to determine the greatest risk
or problems or asset failure.
that could happen in the
future and prescriptive
analytics for recommending
equipment operating conditions and future actons.
With Big Data Analytics for
Predictive Maintenance, the
benefits include increased
asset availability and utilization and reduced operational and
maintenance costs. Example use cases are thermodynamic
equipment such as chillers and mechanical equipment such as
mining trucks.
Use Case 2: Performance Modeling and
Analytics for Predictive Maintenance
The Performance Modeling and Analytics for Predictive Maintenance
solution enables data-driven technologies for performance modeling,
maintenance degradation detection and maintenance effectiveness
estimation.
The solution benefits include early detection of performance
degradation and operational efficiencies and reduction of ineffective
maintenance activities.
Use Case 1:
Big Data Analytics for Predictive Maintenance
The Big Data Analytics for Predictive Maintenance solution is made
up of a common suite of data-driven technologies that provides
descriptive analytics for insights into equipment performance and
maintenance activities, predictive analytics for forecasting failures
Performance Degradation Detection
Maintenance Effectiveness Estimation
Performance
Degradation
Health Index
SOLUTION PROFILE
Hitachi Predictive Maintenance Solutions Enable Increased
Asset Utilization, Availability and Reduced Costs
Time
Percentage
SOLUTION PROFILE
Solution Overview:
■■
■■
Early detection of performance degradation by monitoring the
changes in key performance indicators over time compared to the
ideal performance
Use Case 4:
A Hybrid Approach for Predictive Analytics
Physical Model
Assessment of the effectiveness of maintenance activities and
actions by comparing before and after performance for better
maintenance planning
Simulate
Use Case 3:
Failure Prediction for Predictive Maintenance
Equipment
Event-based Failure Prediction
Sensor-based Failure Prediction
Using Anomaly Detection
'#
(#
)#
#
Sensor Data
#
!&#
!'#
!(#
#
!'#
!$#
!"#
Normal
%'
%&
$
&
&
&#
Normal
clusters
!&#
S23
'#
!(#
!(#
!%#
E71
(#
!)#
!'#
#
Anomaly
Anomaly
Measure
!&#
!%#
!$#
"
!"#
Anomaly data
!
E72
Using Classification Models
Description
Predicted
Event
Impact
Confidence
E71, E72, E83
Electrical System,
Engine, Tires
S23
Standby
99%
E71_A
Engine
S23
Standby
96%
E77
Hydraulic Oil Leak
S23
Standby
95%
E71, E78
Electrical System,
Propulsion
E72
Engine
64%
A challenge companies
are often faced with is
the limited amount of
failure data available for
Statistical Model
training the machine
Field data
Predict
learning model. This
is especially relevant
in regards to reliable
equipment that often do not fail in the field. A physical model can simulate normal and faulty behavior. Once a statistical model is learned
from simulated and field data, it can predict the severity of fault mode
over real time data.
Interested in learning more? Be sure to visit us at:
http://www.hitachi-america.us/rd/about_us/bdl/
Classifier Output
Event
Sensors
E83
Predicted Faults
Solution Overview:
■■
■■
■■
Normal
Fault
The Failure Prediction for Predictive Maintenance solution enables
equipment prediction and component failure by monitoring sensor
and event data. The solution benefits include increasing equipment
availability, avoiding catastrophic failures and reducing repair and
maintenance costs.
(1) Learn association between past
events (2) Use associations to predict
future events
Simulated Data
The Hybrid Approach
for Predictive
Maintenance solution
utilizes the physical
modeling of equipment along with data
analytics for predictive
maintenance.
Event-based failure prediction by learning prediction rules from
historical events and applying the rules over real time event data
Sensor-based failure prediction by learning normal behavior of
sensors and detecting deviation in real time from this normal
behavior as potential failure
Sensor and event based failure prediction by learning classification based models for different categories of failures (from
previous failure instances) and predicting failure by applying
models over real time sensor data
Hitachi America, Ltd
Hitachi America Ltd. Center for Social Innovation
3315 Scott Blvd. 4th Floor
Santa Clara, CA 95054
HITACHI is a trademark or registered trademark of Hitachi, Ltd.
All other trademarks, service marks, and company names are properties of their respective owners.