Predictive Maintenance
Detect malfunctions in
systems and machines,
before they occur.
With Predictive Maintenance, you can detect critical machine conditions in time before actual faults and production downtimes occur. Not only is it possible to predict when and where a fault might occur, but you are also supported in identifying the fault. With this procedure you will receive answers about the condition of your machine, when a failure is to be expected, which component will be affected, why the failure occurs and how you can avoid a failure.
Predictive Maintenance uses historic data of an asset to train a model for statistic analysis and machine learning. These models can be used to predict future behavior of the asset and help identify critical machine states to take countermeasures.
Predictive Maintenance
Benefits at a glance
- Always online
prevent unplanned downtime with component-level defect identification
- High availability
improvement of plant availability and production quality
- Life time
longer service life of machines and equipment
- Efficacy
improve overall equipment effectiveness (OEE)
- Oriented to need
reduce costs (resources) through demand-oriented maintenance
- Rolling
continuous optimization of machines, processes and products
- Schedulable
improved planning of maintenance cycles to keep downtime as short as possible
- Customized
identify advanced use case applications
What is it about?
Basic concept of Predictive Maintenance
Data collection at the production units
A production unit is equipped with a number of sensors and therefore monitored during operation. The resulting measurement data is stored in a database. They provide a complete historical description of the activity of the production unit and its environmental conditions. In addition, all occurring failure and maintenance events are recorded.
Data analysis
A mathematical model description of the data is generated by a computer-based analysis of the available data set. Special machine learning algorithms are used, which can automatically detect the correlations and patterns contained in the measurement data and map them with regard to the analysed faults.
Prediction
By applying the learned model, a prediction for the future occurrence of errors can be made on the basis of the currently recorded measurement data. In the probability of critical operating states occurring, a warning can be issued in good time. This knowledge then flows into the planning and implementation of the next maintenance measure. In this way, components requiring maintenance can be replaced before they can fail, thus endangering the entire production process.
Predictive Maintenance procedure
- Business Understanding
Understanding the question and use case in order to define the goal
- Data Understanding
Identification and exploratory analysis of sensor and disturbance data
- Data Preparation
Preparation and cleansing of sensor and disturbance data
- Deployment
Implementation of the created models into the IT infrastructure and generation of alarm messages
- Data Modeling
Use of machine learning or other data mining algorithms
- Evaluation
Verification of the prediction probability of disturbances & optimizations of the model parameters
Phases of our approach
Predictive Maintenance
in your company
1
Workshop
Initial workshop to identify the problem, use case and specification of requirements, processes and architectures as a basis for predictive maintenance.
2
Conception
Conception and modeling with prototypical implementation and subsequent evaluation of the method.
3
Integration
Integration der Lösung in Ihre IT-Landschaft und Einführung in das Unternehmen, gegebenenfalls Schulung Ihrer Mitarbeiter und weiterführende Projektbegleitung im laufenden Betrieb.
Predictive Maintenance in your company
We support you in this!
We support you with a team of experienced data scientists and consultants. From planning, development and implementation to go-live and optimizing your predictive maintenance solution.
When introducing predictive maintenance, our three-step approach has proven its worth. The length of the individual phases depends on your actual need and can therefore be designed according to your individual needs.
