REDUCE PRODUCTION DOWNTIMES
Every year, significant financial losses accrue due to the failure of technology. Small causes often have huge effects in this context because production environments optimised for efficiency and productivity are largely composed of closely intertwined production steps. Unexpected machine or component failures can bring an entire production line to a grinding halt. The essential lesson here is that the costs of production downtimes are generally far greater than the costs of the actual troubleshooting.
PREDICTIVE MAINTENANCE AND QUALITY CONTROLThere are additional, indirect problems of downtimes, e.g. high contract penalties for the breach of contractually agreed service levels such as agreed delivery dates. Today, offering Service Level Agreements (SLAs) are often a must in order to be competitive at all. Some business models are expressly designed to comply with certain service levels, e.g. by limiting machine downtimes to a minimum. In those settings, a thorough monitoring of Service Levels and regular preventive maintenance are crucial for suppliers.
Why predictive maintenance?
This is how compacer makes predictive maintenance workPredictive maintenance is a method to use data-driven procedures to prevent the problems outlined above. Our solution can be used in different areas, but it is basically used wherever machines are operated. compacer will already support you during the first step of data generation:
The type and scope of data analysed for forecasting maintenance works depends on the actual use case. Most machines and systems already generate their own data, e.g. via Programmable Logic Controllers (PLC). Most modern machines already come equipped with an array of sensors. However, additional sensors can be installed at any time to gain further insight. compacer’s integration platform edbic integrates machines and sensor data, analyses them in real time according to definable rules and stores them according to various strategies. If desired, the data contained in the historical memory can be fed to all kinds of analytical systems. The data can then be automatically evaluated by machine learning and statistical methods.
Existing and required data sources are connected via edbic, where all data will be integrated. In many cases, ETL (Extract Transform Load) processes will set up and manage an asynchronous analytics database, the so-called data warehouse.
In order to ensure a complete and error-free data management, edpem will monitor the processes, log any events and alerts an operator if necessary. To ensure a particularly high level of data quality, we can also set up tools for data cleansing.
In the case of real-time and operational intelligence, edbic will process all data directly, e.g. machine sensor data in manufacturing. Often, this will be realised via synchronous data streaming, where the data will be processed instantly upon creation. edbic will still completely monitor all data streams.
Predictive analytics uses data mining in tandem with neural networks and statistical methods, such as CHAID, to develop models that provide a prediction based on the current data situation. We will develop all necessary evaluations and analysis surfaces based on business intelligence tools made by IBM (Cognos and SPSS).
EQUIPPING MACHINES WITH SENSORS
We are flexible and independent, so we can always work with the sensors of your choice. However, we can also recommend a trusted partner if your decision is still open. Rest assured that we will always support you in the area of sensors and that your vendor independence is important to us!
We know all about networking. It will thus be an everyday task for our experts to connect your machines to the existing IT infrastructure.
INTEGRATING DATA STREAMS
Integration is our core business. We effortlessly integrate data streams from sensors and machines.
MONITORING PROCESS AND DATA STREAMS
Gain more transparency in your process and data streams – with our monitoring tool edpem! Detect predefined and unexpected events and react immediately thanks to our sophisticated alerting functions.
MIGRATING YOUR DATA
We migrate your data to Big Data processing environments based on SQL or NoSQL databases or even to server clusters like Hadoop.
BUSINESS INTELLIGENCE AND DATA MINING
In the area of business intelligence and data mining, we collaborate very closely with IBM (Cognos and SPSS), but also leverage additional cutting-edge third party tools.
INTELLIGENCE THROUGH MACHINE LEARNING
We use machine learning techniques to teach smart AI systems to apply their knowledge and improve it through experience.
Support for your
Different analysis options
Whether your data will be analysed by our own solution edbic, a well-known IBM analytics system, or by your proprietary system, we will support you in all three variants.
such as temperature or pressure. These can first be
aggregated, linked or derived to concentrate large
amounts of data into actionable information or to
provide an enhanced quality of information.
aggregated indicators and historical statistics based on collected
machine data to monitor and assess the current state of all
integrate machine data, but also to include logistics data or data on individual work pieces, for example via RFID recognition.
This is especially important where complex production chains need to operate in fine-tuned, interwoven steps and you need to
recognise both malfunctions within individual machines and failures affecting the entire process chain.
that will be compared against the actual sensor data in real time.
This allows you to detect a gradual increase in temperature or
other anomalies early on.
analytics system will learn to recognise as such. But machine
learning technologies can identify patterns that remain invisible
to the naked eye. If any anomalies crop up within in the data,
the system will detect and assess them.
predicting future failures instead.
failures based on threshold values and prior experience.
For example, it is possible to detect cracks in components
through a change in the frequency spectrum
of structure-borne sound.
statistical methods and used to calculate the probability of faults
and failures and when they will occur. This is usually achieved using
prediction models – artificial intelligence, that is – that have been
specifically trained for a particular use case.
the system, i.e. when they have a high probability of failing within the next maintenance interval. This prevents unnecessary costs due
to unplanned plant-wide downtimes.
immediately launch appropriate actions, such as sending alarms
via e-mail/text messages or initiating non-time-critical
processes to prevent future failures and keep downtimes to a
minimum. For example, it can be used in conjunction with
indicators to modify the operation of a machine in such a way that
it produces less waste.