Condition Monitoring and Predictive Maintenance

Operate Your Special Equipment More Efficiently With Predictive Condition Monitoring

With our research and project experience, we develop customized solutions for predictive and condition-based monitoring of individual special purpose machines and test benches in maintenance, their production and quality assurance.
 

Our Mission

  • To make the current wear stock with impact on quality, reliability and energy consumption, measurable and predictable.
  • Digitization of inspections to collect key performance indicators (KPIs) and plans in maintenance and production in an automated way.
  • Monitoring of multiple assemblies with just a few sensors, e.g. on the drive train: motor, inverter, bearing, gearbox, shaft or on machine tools: spindle, tool suspension, workpiece suspension.
  • Multisensory signal analysis with intelligent algorithms that take into account setup and operating states of the system.
  • Incremental model construction to perform TARGET/ACTUAL comparisons of plant states and to clearly diagnose and localize faults.
  • Identification of efficient manipulated variables to operate plants more sustainably (wear, quality, throughput, energy consumption, scrap) or more efficiently.
Our concept allows a step-by-step introduction of condition-based and predictive maintenance.
© Fraunhofer ITWM
Our concept allows a step-by-step introduction of condition-based and predictive maintenance.

Monitoring of the Operating States of Special Equipment

Condition Monitoring – What Is the Condition of My Plant?

Every plant is different and produces individual products. Based on a flexible software framework developed by us at Fraunhofer ITWM, we realize a monitoring system suitable for your plant – consisting of sensors, measurement technology and IPC, which calculates the current conditions of the plant, visualizes them as characteristic values and makes them available to production management systems. You can expect the following from us:

Mathematics and Technology for Monitoring Process Flows

We help to connect existing sensors and to install additional sensors to monitor process variables (e.g. acceleration, temperature, torque, pressure, speed) of the plant sufficiently accurately.

Überwachung von Getriebespektren im SOLL-IST-Vergleich
© Fraunhofer ITWM
Eine einfache Überwachung von Getrieben erfolgt über einen SOLL-IST-Vergleich von gemessenen Frequenzspektren.

Our approach includes mathematical methods to capture as many assemblies with as few sensors as possible. We evaluate the information content of acquired data in terms of sensor, system and process noise and filter for patterns with relevant information. To be just as time-efficient, we conduct a »design-of-experiments« with you, if necessary, so that we specifically collect data on different operating points. In addition, we let you monitor components that are inaccessible to sensors through virtual sensors and link maintenance logs, recipes and other information from production and maintenance management to include in monitoring, for example, the following:

  • Kinematic frequencies of components (e.g. shafts, bearings, gears, drives, inverters)
  • Influences due to maintenance activities (of the last lubrication)
  • Control variables of the production (SPS)

Automated Collection of KPIs on Maintenance, Production and Energy

We create a monitoring system for the supervision of the plants with characteristic values from maintenance, production and energy, which represent the operating conditions with physically interpretable sizes suitable to the current operating points (e.g. the manufactured product). In addition, we identify the effects of occurring production events and help to identify anomalies at an early stage.

Diagnosis and Evaluation of Plant Conditions

Condition-Based Maintenance – Why Does the Current Condition Deviate From the Target Condition?

States of the same plant differ strongly when different products are manufactured. We at Fraunhofer ITWM develop a diagnostic system, which classifies the monitored states into categories according to the process. Thus, we provide condition-oriented hints for maintenance and repair measures to be planned. Here you can expect from us.

AI Procedures Adapt Diagnostic System to Your Processes

We evaluate the monitored condition of the plant with categories to provide indications for root cause analysis and maintenance or repair plans. Faults are identified and localized more quickly.

The categories of the diagnostic system match the process and relate monitored parameters of component condition of the components (e.g. shafts, bearings, gearboxes, drives, converters) to relevant target variables (probability of failure, safety, quality or scrap, energy consumption). This enables you to make needs-oriented decisions for your maintenance measures.

Diagnose Predictive Maintenance
© Fraunhofer ITWM
Eine Fehlerdiagnose eines Antriebstrangs wurde mit simulativen Schadensbildern, einer sensorischer Spektralüberwachung und gemessenen Schadensfrequenzen realisiert.

Using machine learning, we train suitable classification models for your process to enable a simple, technical and qualitative assessment of the condition for plant managers and technicians. The trained models run either in your control cabinet on industrial PCs or in the cloud in the case of remote service platforms. Training of the models for new states is usually done on a suitable server in IT.

Predictive Maintenance – When Can the Eqiupment No Longer Guarantee Reliability or Quality?

Trends in the change of plant conditions provide information about the remaining wear stock. We at Fraunhofer ITWM develop predictive maintenance systems to forecast the remaining operating time until the next necessary maintenance or servicing measure. 

This predictive maintenance is considered the optimal approach for efficient planning of maintenance windows, spare parts storage, and warranties. Here's what you can expect from us.

AI Methods Adapt Their Forecasts to Your Processes

We detect the trend in the change over time (degradation due to wear) of the condition and characterize this trend using suitable machine learning methods. In doing so, we take into account different loads that the plant has experienced since the last maintenance. The resulting model allows a simulation of the machine's condition and predicts the time until a threshold is reached at which operation is judged to be too risky. The trained models run either in your control cabinet on industrial PCs or in the cloud in the case of remote service platforms. Training the models for newly added states is usually done on a suitable server in IT.

Video: Predictive Maintenance – Forecast of the Differential Pressure

Privacy warning

With the click on the play button an external video from www.youtube.com is loaded and started. Your data is possible transferred and stored to third party. Do not start the video if you disagree. Find more about the youtube privacy statement under the following link: https://policies.google.com/privacy

Predicting differential pressure in a wood pellet-fueled gasifier enables predictive maintenance from the last cleaning event.

Optimization and Adaptation of the Operating Condition

Cycle of Optimization
© Fraunhofer ITWM
Cycle of Optimization

Predictive Control – How Do I Optimize the Operation in Production on This Plant?

The condition-based monitoring of plants can be used in many ways. At Fraunhofer ITWM, we use recorded and predicted operating states of your plant and evaluate the sensitivity with respect to applied control inputs in order to optimize them as further target parameters in production control. You can expect the following from us:


Condition-Based Maintenance Enables Automation With Digital Twins

We extend the automation of your plant by converting and using the trained models via the plant condition to find efficient manipulated variables for targeted operating points in the current operating state. This corresponds in essence to the idea of a digital twin via the plant state. The inverted models run in on-premise installations in the control cabinet on industrial PCs. These are connected to the PLC controllers of your plant via additional manipulated variables.

Set-up Your Test Benches for Digitalization of Production

Profiles – How Do New and Used Products Behave in Different Conditions?

If we collect the states for both manufactured and returned components and machines, we obtain an empirical database on good and »bad« states, due to wear and defects. At Fraunhofer ITWM, we build and expand your test benches so that we can evaluate and predict these qualitative profiles about states with life estimates. This enables you to develop digital business models regarding warranty, leasing, service, or spare parts management. Here's what you can expect from us in the process.

We support you in profiling by updating test and inspection benches from manufacturing and quality assurance, both sensor-based and through software, in order to record and store condition profiles in a structured manner as a kind of fingerprint for newly manufactured machines and for returns.

Contectless Torque Acquisition
© Fraunhofer ITWM
Our systems are supplied with an inductive sensor for contactless torque measurement.

We assist in structured acquisition by performing experiments in a controlled and, if possible, automated manner via efficient design-of-experiment (DoE) approaches. The collected profiles serve as references for quality-assured good states and failure patterns for damage states. On this basis, threshold values for condition-based or predictive maintenance can be empirically determined and optimized.

We extend your test bench with sensors and an industrial PC to record condition profiles, store them in a database and evaluate them. We integrate the solution into your existing IT infrastructure or install our own databases and dashboard solutions as required.

Start Faster With Vendor-Independent Software

Software-Tool Predictive Maintenance
© Fraunhofer ITWM
Wir stellen eine vom Messsystem unabhängige Software bereit, um passgenaue Instandhaltungsanwendungen auf Basis von Signalerfassung und -verarbeitung zur Überwachung und Prognose von Maschinenzuständen zu implementieren.

Profiler – The Integrable Software Framework for Monitoring Machines of All Types

After the decision to digitalize maintenance and production in a condition-oriented manner, the first results should be installed in the project as quickly and sustainably as possible. 

With our project experience, we at the Fraunhofer ITWM provide a software framework (Profiler) that is independent of sensor and IPC manufacturers and can be used to extend existing systems with the latest mathematical methods of condition-based maintenance. This is what you can expect from us:

The Profiler makes it possible to invest project time as efficiently as possible in evaluations instead of connecting measurement technology.

The profiler bundles the channels of connected sensors and measuring amplifiers via common digital transmission protocols and makes them available in the internal plant network via Ethernet as simple web sockets. Evaluation units can subscribe to one or more channels on the same hardware or other IPCs in the network and execute signal processing, classification, prediction or control algorithms appropriate to the process. The results are again made available as web sockets. The Profiler provides the software basis for monitoring the states of individual plants as well as for diagnostics, evaluation, prognosis, and optimization. Relevant characteristic values, evaluations and prognoses are transmitted via clients that transfer results to databases, dashboards, files, oscilloscopes and the like.

Requirements

  • Operating system (Windows, Linux)
  • x86 or ARM
  • Ethernet interface
  • One core per sensor
  • Memory and CPU vary depending on algorithm and data rate