Process Control and Predictive Maintenance Methods

Process Control for Printing Technology

Process optimization with the help of an inline feedback loop. Process data and images from the inline AOI system are used as input.
© Fraunhofer ISE
Process optimization with the help of an inline feedback loop. Process data and images from the inline AOI system are used as input.
Digital twin of the printing form: any mesh configuration can be mapped and used for optimization calculations in advance.
© Fraunhofer ISE
Digital twin of the printing form: any mesh configuration can be mapped and used for optimization calculations in advance.

Increasing technological demands on the metallization of Si solar cells require new approaches for optimizing the screen printing process. Complex parameter interactions must be understood and taken into account in order to make targeted process fine tuning. Since the parameter space has a large number of individual variables, which in most cases are interrelated, optimizing the parameters with sufficient effect is not trivial. In order to incorporate the continuous development of the process components into the approach, Fraunhofer ISE is working on an automated solution in digital space that can determine and correct any changes and their influences in the parameter space. In this context, artificial intelligence, for example, can serve as an effective tool for recognizing patterns and making coordinated adjustments in the parameter space.

As printing tool components, such as printing forms, metal meshes or metallization pastes, are subject to changes such as warping or drying out during the printing process, there is a need to dynamically adjust the process parameters depending on the specific number of cycles. An inline feedback loop, which collects information from the printing line's optical inspection (AOI module), is designed to provide suggestions for real-time parameter adjustments using an integrated AI model. This enables immediate adjustment of the parameters according to the condition of the components. By using machine learning, errors are also detected at an early stage and the process parameters are dynamically adjusted to ensure consistently high quality. 

Our R&D services include:

  • screen optimization:layout-specific adjustment of the mesh parameters using an optimization algorithm
  • quality control (inline): upgrade of existing 2D AOI systems to 3D measurement by means of a simple software update
  • quality control (mesh): specific implementation and customization of AI-based incoming inspection systems for fast and efficient inspection of virgin material (mesh)
  • training and implementation: provision of the necessary know-how for the integration and use of machine learning solutions in existing production lines (printing technology)

Industry 4.0 for Wet Chemical Processing

Process control and predictive maintenance: temperature monitoring
© Fraunhofer ISE
Temperature monitoring with simple limit value monitoring (top), model-based with fixed tolerance range (middle), situation-adapted as probabilistic model with automatically generated tolerance range.
Process control and predictive maintenance: data cockpit
© Fraunhofer ISE
Data cockpit of the labflux time series database for recording sensor and actuator data in an alkaline etching bath during the production of silicon solar cells

Data transfer and databases: A system-specific data evaluation requires sensor and actuator data in a well-structured database. The data are usually stored as time series. In addition to the data from the system's own sensors and actuators, other data sources that either record the environmental conditions or are installed within retrofits can also be important.

In our Industry 4.0 model system for wet chemical processing, we read data from the equipment system control as well as other sensors and IoT devices. Data are transferred via MQTT, which enables flexible and secure distribution of the data to various databases. The data are stored and aggregated in our LabFlux time-series database and is available for further analysis.

Data evaluation and predictive maintenance: In addition to monitoring the quality of manufactured products, monitoring the processes and the process equipment is also extremely important. Defects that remain undetected over a longer period of time can lead to a gradual loss of quality as well as serious technical faults, which can endanger life and limb. Limit value monitoring alone is not sufficient for reliable identification of system faults. Fraunhofer ISE has therefore developed a concept for fault identification and root cause analysis.

Faults are identified using a physical model or an AI model, which calculates the current sensor values from the actuator settings of a past time window and compares them with the real values. Probabilistic models also enable the calculation of a confidence interval and thus allow a statement to be made as to whether a deviation is of a statistical nature or is due to a system defect. Identified errors are processed in a multilevel flow model in order to draw conclusions about potentially defective components from the errors found.

Our R&D services include:

  • structure of the architecture for data acquisition
  • design and implementation of data streams and databases
  • development of digital models of the process equipment and modules for automated fault identification in the process equipment
  • development of digital models for root cause analysis

Further Information on this Topic

Research Topic

Artificial Intelligence and Data Management

Business Area

Photovoltaics:​ Production Technology and Transfer