Guarantee availability in production
Avoid downtimes, reduce maintenance costs and ensure maximum machine utilization with our predictive maintenance solution.
Your advantages at a glance
Example hydraulic cooling and filtration system:
Precise predictions
Monitoring of over
43,000 sensor data per
cycle
Increased efficiency
Targeted maintenance
reduces unplanned
downtime and costs.
Real-time transparency
Clear forecasts for
critical hydraulic
components.
Flexibility
Scalable for individual
machines or entire
fleets.
Technical setup
Our predictive maintenance solution is based on the Azure Cloud and the Telefónica IoT Hub. It can be transferred to other cloud providers such as AWS or Google if desired.
System structure
To illustrate the functionality, we provide an example application with sample data from a hydraulic system for cooling and filtration. The data is real data. This gives you a profound impression of what a predictive maintenance solution could look like in real operation.
The interfaces, evaluations and automatic actions of the demo app can be configured individually and, in our case, correspond to the functionality of the example system.
Our example systems
Our hydraulic system for cooling and filtration (example) consists of the following components:
- 4 hydraulic cooling and
filtration units (machines) - 7 sensors per machine (pressure,
motor power, mass flow,
temperature, vibration) - 60 to 6,000 measured values per sensor per minute (depending on sensor type)
Our maintenance forecasts
From this we derive 4 maintenance-relevant target values; they are the result of our prediction based on the
input values of the sensors and trigger maintenance messages:
- Cooling (% efficiency)
- Valve tightness (% efficiency)
- Pump leakage (probability)
- Hydraulic compressor pressure (bar)
If a target value deviates from the expected interval, a warning is issued, which can automatically trigger subsequent processes
or at least describes the expected status of the machine in the PdM dashboard.
Maintenance measures can then be carried out proactively.
IT system structure
Our sample application uses the following state-of-the-art technologies.
Telefónica IoT Hub KITE
collects all data from the sensors
via IoT 5G, WLAN, LAN or
other network connections.
On request, Telefónica
can also supply IoT hardware
(microcontrollers) from leading
manufacturers.
Azure Data Lake Storage
This is where the IoT data is
is stored. Alternatively
also be stored in the Telefónica IoT Hub
KITE, so that
the Azure Data Lake is optional.
Azure Databricks
Databricks prepares the data
data. Here they are normalized,
checked for errors, supplemented and
and checked for plausibility.
Azure Machine Learning (ML)
provides the model environment for
our machine learning models,
that predict how sensor data
influence the tightness of valves
influence the tightness of valves. This is where the
mathematical prediction.
Azure Kubernetes
Services (AKS)
The Azure ML model runs
within Kubernetes,
so that it can be operated
scalable operation
scalable. It represents the computing unit
unit.
Business Automatica Apps
The "PdM Dashboard" is the user
user interface of our
sample application that displays the
prediction results
and automations
the data of which can be
systems such as PLS, ERP, TMS or
BI.
Advantages of this setup
Granularity of the data
High measurement frequency provides in-depth insights into short-term fluctuations and anomalies
Holistic monitoring
Different sensor types cover all critical states of the hydraulic systems.
Precise forecasts
Thanks to the processing of over 43,000 data points per cycle, you receive reliable predictions and in-depth insights into short-term fluctuations and anomalies
Discover the benefits of data-supported maintenance.
Try out our predictive maintenance solution with no obligation.
Predictive maintenance - turning data into decisions.
How the demo app works
How it works: The technology behind it
Our predictive maintenance solution combines state-of-the-art machine learning algorithms and practical
applications to create a data-based foundation for maximum machine availability.
Machine learning models
- Algorithms: XGBoost, specially optimized for
tabular data with high
dimensionality.
- Hyperparameter tuning: Use of Bayesian Optimization for the optimization of learning rate, tree depth and other key parameters.
- Forecast for each destination: Separate models for cooler, valves, pump and accumulator.
Process automation
- Data acquisition: real-time integration of sensor data into the platform.
- Pre-processingNormalization and merging of the data.
- Analysis: Anomaly detection and
predictions with trained models.
- OutputResults in JSON format with machine-related parameters (e.g. timestamp, machine number).
Contact us now
Don't wait until the next machine breakdown brings your production to a standstill.
FAQ
What does our PdM solution do?
Our PdM solution enables the early detection of faults and maintenance issues on machines and systems before problems occur. We offer an end-to-end solution that can be integrated into existing machines and systems so that the investment is kept to a minimum and the savings are immediate.
Who are the target customers of our PdM solution?
On the one hand, medium-sized machine and plant manufacturers who want to make their spare parts business more predictable and be perceived as more reliable and of better quality in the eyes of their customers. On the other hand, companies as users of machines and industrial plants that want to avoid production interruptions and reduce their maintenance costs through better utilization of components.
What makes our PdM solution better than other providers?
It is 1. cheaper because we only make additions where necessary, 2. requires fewer changes to existing systems, 3. interacts with all sensors and IT systems, 4. is open to all technologies, although we first developed it on Azure, and is structured so flexibly that every customer situation can be served with manageable effort, costs and duration.
What are suitable steps towards a PdM solution?
The functional, i.e. technical, understanding of the customer's desired target situation is crucial: 1. which machine or system malfunctions should be detected and predicted with the help of PdM? 2. has the correlation between measured values (sensor technology) and error states (failures, maintenance indication, drop in performance, etc.) already been recorded numerically? 3. if not, then we can provide support here. 3. which sensors are used? If none, we can help you select suitable sensors.
What special requests can be fulfilled?
Basically all of them. As a rule, the supply of process control systems or special evaluation solutions (control towers, dashboards, etc.) is required or their development is requested. We fulfill these requests directly, as we control all stages of our PdM solution and do not depend on the functionality of third parties.
Do we also supply sensors?
No. However, we arrange for partners such as Vega, Endress & Hauser, IMF and many others to supply, install and commission them. We coordinate all related activities.
What does a PdM introduction cost?
The minimum project costs are EUR 15,000, provided the sensor technology is already available and the systems to be connected have interfaces (e.g. Wonderware, SAP, Rockwell, etc.). If hardware has to be procured or more complex developments are required, the costs increase. However, we calculate this in the quotation phase.
What running costs can be expected?
We offer a consumption-based license model. We can estimate the running costs based on the amount of expected measurement and forecast values and the number of machines/systems. We strive to minimize operating costs thanks to automation.
You can rely on our expertise in predictive maintenance
maintenance to maximize the reliability of your systems and
and increase your competitiveness.
COMPANY
Business Automatica GmbH
Eisenbahnstraße 28
67725 Börrstadt
Germany
CONTACT
Gernot Reinmüller
(Managing Director)
Phone +49 176 3208 3776
Directions via Google Maps