Predictive Maintenance

A technical installation is then reliably available if it is repaired in a timely manner, i.e. maintenance and repair are done directly. This will only succeed if the company can access appropriate resources. The appropriate skilled workers, spare parts and logistics are present.

If systems are operated without repair due to an unidentified need for maintenance of a component, other components are affected. The subsequent maintenance takes longer and becomes more expensive.

Fast reactions at the right time require reliable detection and prediction of damage events in order to provide the necessary resources as required. The goal is a maintenance strategy that detects faults before they occur and allows optimal maintenance to be planned.

© ITWM
Prognosis of the remaining useful life time of a plant by a COX regression

What is Predictive Maintenance?

A reliable prediction of future events is an integral part of any predictive maintenance system. An important key lies in the patterns of past events. The recognition of complex event patterns as well as their own dynamics and trends are the playing field of the system we developed.

According to DIN 3105, predictive maintenance is an extension of condition-based maintenance that predicts when the system will failure . The main goal here is to make the best possible use of the component's stock and, at the same time, to rectify any faults or failures before they occur.

© ITWM
Classification of failure events in an information space optimized by PCA

Five reasons for predictive maintenance

Predictive maintenance optimizes service intervals by determining the individual operating lifetimes. This maximizes the useful lifetime of systems, as maintenance costs are planned in sufficient time to meet predicted demand.

Early detection of anomalies in the plant or its performance.
Identification of root causes of unplanned failures / faults
Planable estimation of the remaining useful life of plants and components.
Reliable prediction of failure in the near future.
Efficient planning of upcoming maintenance actions.

Models and techniques

Depending on the objective, statistical or stochastic models recognize, classify, predict and even control the different states of your plant. Below are a few examples:

FIELD	DesCriptivE	DiagnostiCS	PrEdiCtivE	PrEsCriptivE
QUESTION	WHAT is the operating state?	WHY did the error occur?	WHEN does the next error occur?	HOW will the operation be continued?
Information	Data acquisition and pre-processing Feature extraction
use cases	Detect and assess operating conditions anomalies events	Classify anomalies events	Predict trends life-time events	Control maintenance time plant operation
Selection of our algorithms	Detect and assess signal analyses threshold analyses self-organizing maps	Classify deep learning decision trees Bayesian networks	Predict mixed-effect models event models joint modeling	Control predictive control reinforcement learning

Our portfolio and range of services

Predictive maintenance can be introduced quickly, especially where a wide range of data is already being collected. However, it will also be worthwhile retrofitting your plants with monitoring systems.

	Our Offer
Machinery and equipment	Sensor selection and placement Virtual sensors Condition monitoring System modeling and simulation with digital twins
CommuniCation	Integration into common IOT platforms, communication protocols and networks
DatA MANAGEMENT	Efficient data collection through design-of-experiments System and event models Data cubes
Forecasting and evaluation	Identification, implementation and application of suitable machine learning algorithms Diagnosis and prognosis of data streams Analytics, reports and dashboards
ProCess integration	Reliable control of connected systems Optimization of maintenance and service Enterprise resource and lifecycle planning

Fraunhofer Institute for Industrial Mathematics ITWM