Solution overview

Cisco public

SparkPredict Anomaly
Management on Cisco
HyperFlex Edge
Reimagine industrial maintenance with AI.

Contents
2 Overview
3 The need to evolve maintenance programs
3 How it works
3 Main capabilities
3 Power your predictive maintenance with
Cisco HyperFlex Edge and 2nd Generation
Intel® Xeon® Scalable processors
3 Sizing on-premises computing resources
for SparkPredict: Base deployment
requirements
4 Use cases
4 Cisco Capital financing to help you achieve
your objectives
5 Success stories
5 Cisco Toronto Innovation Center
5 A member of Intel AI Builders
6 The Cisco Advantage
© 2020 Cisco and/or its affiliates. All rights reserved. 6 Next step
Solution overview
Cisco public

Benefits The SparkPredict® software from SparkCognition unlocks the value of industrial
sensor data, allowing you to predict and prevent unique catastrophic events, improve
The SparkPredict software provides these
operational efficiency, and optimize scheduling and maintenance. Powered by Cisco
benefits to your organization: HyperFlex Edge and optimized for 2nd Generation Intel® Xeon® Scalable processors,
this is a predictive maintenance solution that’s easy to deploy and seamless to manage.
• Increase competitive advantage: Meet or
exceed production targets while reducing
unnecessary maintenance costs Overview
• Codify knowledge and operations: Process industries lose billions of dollars to unplanned stoppages caused by undetected and
Continuous learning as assets unexpected mechanical equipment failure. With a SparkPredict and Cisco HyperFlex Edge solution you
age and share subject matter can use your existing sensor data to make accurate and actionable predictions of when a machine will
expertise across the team break down so that you can resolve issues before they disrupt production and cause losses.

• Increase operating efficiency: SparkPredict predictive analytics software uses sophisticated algorithms to predict asset failures,
Simplify operational processes and reduce maintenance costs, and improve system efficiency. Next-generation prognostics based on
optimize resource utilization artificial intelligence (AI) learn from sensor and control system data to better understand the future
behavior of each asset. The algorithms then raise alerts for sub-optimal operations and identify
• Increase operational reliability: Improve impending failures long before they occur.
visibility into asset health at scale
and prevent unplanned downtime

• Accelerate diagnostics and decision-

making: Equip subject matter experts
with context to triage and action most
alarms in less than 10 minutes Adaptable Data Ingestion Asset Modeling, Management, & Execution Insight Extraction & Display

Model Building-as-a-
Model Retraining Engine
IoT Sensor Data

Service Alerting
Data

Action
Ingestion & +
Preparation Insight
RT Data pipeline from
Pipelines Est. Model Run-Time Engine & Expert Input
models
Execution

Figure 1. SparkPredict: Built to Deploy at Scale

Solution overview
Cisco public

The need to evolve maintenance programs Power your predictive maintenance with
Maintenance programs need to evolve. With an accelerated rate of Cisco HyperFlex Edge and 2nd Generation
operational change across industrial sectors, emerging gaps in subject
matter expertise, and increasing financial pressures, there is no longer room Intel® Xeon® Scalable processors
for costly unscheduled downtime, unexpected failures that may put workers Cisco HyperFlex Edge and Cisco Intersight management make deploying at a
at risk, or inefficient operations. remote location as simple as connecting power and the network. Edge nodes
connect to the Cisco Intersight platform through a secure Transport Layer
Traditional maintenance programs have a value ceiling. Systems such as
Security (TLS) connection, allowing you to manage your cluster from any
visual or instrument-based inspection and real-time condition monitoring
location. SparkPredict’s algorithms are optimized to run on 2nd Generation
generate too many false positive alerts, do not enable timely decisions, are
Intel® Xeon® Scalable processors in your Cisco HyperFlex Edge nodes.
difficult to adapt and modify, and are highly dependent on human expertise.
But sensor data and AI, used together, offer the opportunity to evolve your
maintenance and analytics model. Sizing on-premises computing resources for
SparkPredict: Base deployment requirements
How it works
SparkPredict® software is a predictive maintenance and anomaly Server vCPU/Core Memory Disk
management product. SparkPredict uses automated model building
techniques that employ supervised, unsupervised, and deep learning Ingestion 8 24GB 1.1TB
algorithms to reduce scaled deployment from years to months. SparkPredict Message queue (up to 5) 2 8GB 1.25TB
takes a customer-first approach and engages with the client to identify the
high-value business objectives and the appropriate level of modeling that Pipeline 16 60GB 600GB
needs to be performed. SparkPredict can then employ all the tags (as many
Database 8 24GB 600GB
as a thousand in some cases) that are needed to model a complex asset, a
component, a failure mode, a process, or a subsystem. With this approach Web 8 16GB 1.1TB
users get a solution designed for their business and their assets. Table 1. Sizing guidelines

Main capabilities
SparkPredict is server based and can be run on-premises, or in a private
or public cloud. The goal is an interface that allows a user to perform a
10-minute triage. An alert is generated, and the features contributing to the
alert are ranked. The user can delve deeper into the alert and features and
see previous alerts that were similar. The user can then quickly identify the
root cause.
Solution overview
Cisco public

Figure 2 provides an overview of the SparkPredict architecture.

Existing data sources Database node Pipelines node Database node

Asset Model runtime

Data ingestion Configuration/

Application web service API

inventory/

Front-end web service API

meta data administration
Monitoring
Job management

UI front end
Data ingestion

Use case-specific
Pipelines dashboards (snap-ins)
Maintenance
logs Data clean/prep Notification
Default user
Events Dashboards
Data clean/prep Monitored assets
Top contributors
Data Listener
historian similar alerts
Alert labeling
Data
Data ingest configuration Model Alerting Results
DB rules DB DB DB

Figure 2 SparkPredict architecture

Use cases
Table 2 describes some typical use cases for the SparkPredict solution.

Industry name Description

Oil and gas • Provide advance warning of both known and unknown asset failures or outages.

Energy • Provide insight into asset failures with anomaly detection.

• Classify asset operating states and normal behavior.
Manufacturing
• Improve operational efficiency and safety.
• Plan optimal maintenance scheduling and distribution of parts.

Table 2 Use cases

Solution overview
Cisco public

Success stories Cisco Toronto Innovation A member of Intel AI Builders

Improving offshore platform production with AI
Center The Intel AI Builders Program is an ecosystem of
industry-leading software and hardware partners
Digital business is driving transformation
Challenge: A major oil and gas company with who have a shared mission to accelerate the
and disruption at an unprecedented rate. No
high-volume offshore platforms needed to adoption of artificial intelligence across Intel
industries will be exempt. Organizations need
increase production potential and improve overall platforms. SparkPredict’s predictive algorithms
a trusted partner with the expertise to build a
platform safety. take advantage of Intel’s industry-leading,
smarter, faster, safer business. Cisco is helping
Solution: The company partnered with workload-optimized platform with built-in AI
bring that vision to life.
SparkCognition to deploy AI-powered predictive acceleration, providing the seamless performance
analytics across multiple critical subsystems as At the Cisco® Toronto Innovation Center, our foundation for the data-centric era. The 2nd Gen
well as the company’s fleet to predict impending goal is to help businesses experiment, test, and Intel Xeon Scalable processors enable a new
failures and optimize maintenance activities. co-create solutions to solve real-world business level of consistent, pervasive, and breakthrough
challenges. The Innovation Center, located in performance.
Benefits: By using SparkPredict® AI analytics,
downtown Toronto, is one of 12 state-of-the-art
software, the customer improved production by
global facilities that aims to inspire organizations
1 to 4 percent, or up to US$30 million annually
and showcase digital innovation and development
per platform.
with real use cases. Along with our global
Read the full case study here. ecosystem of partners, startups and researchers
we help businesses solve challenges and thrive
Identification of vane failure in turbine data in an always-on, ever-changing, constantly
connected world by fostering collaboration and
Challenge: After a unique combustion turbine vane
empowering organizations of all types and sizes
failure led to a two-month outage and US$30
to bring concepts to reality.
million in lost opportunities, a combustion turbine
original equipment manufacturer (OEM) wanted to At the Cisco Toronto Innovation Center,
find a new way to prevent such costly outages. we are demonstrating how you can bring
Solution: The SparkPredict® solution with its powerful predictive maintenance capabilities
asset-centric model was evaluated for its ability to any location using Cisco HyperFlex Edge
to predict failures in advance. featuring 2nd Generation Intel® Xeon® Scalable
Benefits: The SparkPredict® software provided processors.
one month’s advance notice of a failure. The To find out more, contact us at
solution is now moving into full deployment. The TOinnovate@cisco.com.
operator expects to reduce costs by 30 percent
using this solution.
Read the full case study here.
Solution overview
Cisco public

The Cisco Next step

Advantage Interpret your valuable data to detect malfunctions and reduce production downtime. With Cisco
HyperFlex Edge powered by 2nd Generation Intel® Xeon® Scalable processors and SparkPredict, you
Cisco HyperFlex Edge provides the reliable, have a powerful solution to recognize failures before they happen and create a more efficient and safer
scalable and easy-to-deploy computing operation.
capabilities that you need for your mission Find out more about the power of Cisco hyperconverged infrastructure at the edge:
critical AI solution running at the edge of
the network. With the simplicity of Cisco • https://www.cisco.com/c/en/us/products/hyperconverged-infrastructure/hyperflex-solutions.html
Intersight remote management and powered • https://www.cisco.com/c/dam/en/us/products/collateral/hyperconverged-infrastructure/hyperflex-hx-
by 2nd Generation Intel® Xeon® Scalable series/hyperconvergence-at-the-edge-aag.pdf
processors – you have secure and flexible
Find out more about SparkPredict:
computing power close to your data.
• https://www.sparkcognition.com/products/sparkpredict/

For more information about Cisco® solutions for AI/ML workloads visit:

• https://www.cisco.com/go/ai-compute

© 2020 Cisco and/or its affiliates. All rights reserved. Cisco and the Cisco logo
are trademarks or registered trademarks of Cisco and/or its affiliates in the U.S.
and other countries. To view a list of Cisco trademarks, go to this URL: www.
cisco.com/go/trademarks. Third-party trademarks mentioned are the property
of their respective owners. The use of the word partner does not imply a
partnership relationship between Cisco and any other company. (1110R)

© Cisco and Intel work together to optimize solutions from the edge, through
the network, to the data center. Cisco’s innovative UCS servers and HyperFlex
hyperconverged solutions, powered by Intel Xeon processors, Intel Optane™
technologies, and Intel Ethernet products, provide high performance with lower
TCO, simplified operations, and a modern, more secure data center. Customers
can benefit from joint, scalable solutions, designed to solve key market
challenges in multiple industries. Together, Cisco and Intel create a trusted
ecosystem for hardware, software, and services that bring intelligence and
analytics to your business.

220167-2 08/20