Skip to main content

Hitachi in Oceania

Anomaly Detection Solutions


There are now far more connected devices than people in the world, with an expected 50 billion connected devices by 2020. This is opening up billions of dollars in global economic opportunities, as industries explore innovative ways to use pervasive connectivity, intelligent machines, machine data and advanced analytics to solve new and existing business challenges.

Big Data is the enabler for Hitachi’s innovative Anomaly Detection Technology solutions approach, and we specifically focus on the Internet of Things that matter to business. We explore practical big data applications and how this information can be used to create transformative solutions for enterprises around the world. But that’s only the beginning. We are also anticipating exponential growth in machine-generated data and big data, and we are building and refining advanced analytics solutions to help enterprises draw deeper, faster insights from their data.

Benefits of Hitachi’s Anomaly Detection System

Imagery Analysis Services Approach

anomaly-detection-solutions-benefits part 1

anomaly-detection-solutions-benefits part 2

anomaly-detection-solutions-benefits part 3

anomaly-detection-solutions-benefits part 4

Application

  • Predictive Failure Analytics
  • Condition Based Maintenance

Services

Conventional Maintenance
Predictive Maintenance
Anomaly Detection Algorithms
Predictive Failure Analytics Approach
Condition-Based Maintenance
Approach to Condition-Based Maintenance

Conventional Maintenance

Preventive maintenance with periodical inspections and parts replacements. Unexpected failures occur between intervals of periodical maintenance.

Conventional Maintenance

Predictive Maintenance

Collects sensor data in real-time and predicts potential failures before machines break down.

Predictive Maintenance

Anomaly Detection Algorithms

The anomaly detection system uses our Vector Quantisation Clustering (VQC) and Local Subspace Classifier (LSC) algorithms.

These perform machine learning on normal-status sensor data and create indication of differences between the data to be monitored and the learned normal data group. The system then evaluates whether the result is normal or abnormal.

Traditional methods such as the Mahalanobis-Taguchi (MT) systems can only be applied when sensor data has a normal distribution. Hitachi’s algorithms are resistance to effects from the data distribution.

Since the algorithms are model-free, they can respond flexibly without the need for model construction or simulations for each status change, even when there is a major change in a device or system operation status.

The optimum system configuration uses each algorithm separately according to the device or system to be monitored, or to the characteristics of the abnormality to be detected. A drawback of conventional data mining functions is that causes are difficult to explain when diagnosis results are derived from complex sensor signal correlations.

Therefore, the Hitachi system has been designed to simplify the cause analysis by outputting an ordered lists of the sensor signals responsible for a detected status change.

VECTOR QUANTISATION CLUSTERING (VQC)

Vector Quantisation Clustering (VQC)

LOCAL SUB-SPACE CLASSIFIER (LSC)

Local Sub-Space Classifier (LSC)

Predictive Failure Analytics Approach

  • Machine data is automatically diagnosed by Hitachi’s data mining technology
  • The failure signs are then reported to the maintenance service division

Predictive Failure Analytics Approach Part 1

Predictive Failure Analytics Approach Part 2

Condition-Based Maintenance

Why ‘Condition-Based Maintenance’?

  • Realise an optimum maintenance plan based on a degradation level
  • Minimise maintenance duration and increase operating duration
  • Avoid unplanned shutdown by real-time monitoring and diagnosis
  • Visualise and share the needs of maintenance by quantifying degradation level

A MAJOR CHALLENGE OF ASSET OPERATION IS TO BALANCE RELIABILITY AND AVAILABILITY

Reliability And AvailabilityCondition-Based Maintenance - Reliability And Availability

Approach to Condition-Based Maintenance

Condition-based maintenance is implemented using a combination of condition diagnosis and risk evaluation, leading to effective maintenance without damage to equipment or over-maintenance.

Traditional Maintenance

CORRECTIVE MAINTENANCE

  • Maintenance after trouble occurred
  • High risk of damage or unplanned shutdown caused by degradation

TIME-BASED MAINTENANCE

  • Periodical maintenance by pre-evaluated lifetime of parts
  • Maintenance costs is high as condition of parts have no relation

Predictive Failure Analytics Approach Part 2

Condition-Based Maintenance

Effective maintenance based on the degradation degree of components. Realise high reliability by avoiding damage to equipment and over-maintenance.

  • MAXIMISATION OF AVAILABLITY
  • MAINTENANCE SUITABLITY

THE FOLLOWING STEPS ARE REQUIRED TO ACHIEVE CONDITION-BASED MAINTENANCE

Steps to Achieve Condition-based Maintenance

Steps to Achieve Condition-based Maintenance part 1

Steps to Achieve Condition-based Maintenance part 2

Steps to Achieve Condition-based Maintenance part 3

Case Study

Electronics Manufacturing Machinery

Electronics Manufacturing Machinery

Damage of feed bar causes unscheduled stoppage of printing press.

Gas Engine Maintenance

Gas Engine Maintenance

Advanced predictive diagnosis technology has been applied to condition based maintenance system for gas engine generators on a commercial basis.

Corporate / Office Profile

Corporate / Office Profile
Corporate / Office Name Hitachi Australia Pty Ltd
Address Suite 801, Level 8, 123, Epping Road, North Ryde NSW 2113 Australia
Tel +61 2 9888 4100 or 1800 HITACHI
Fax +61 2 9888 4188