For the quality assurance in industrial processes the monitoring of colors is an important factor.
Production parts are not meant to show drifts in their coloring. At the moment the monitoring concerning this issue only takes place on a sample basis. However, to effectively control such processes and to be able to react immediately when errors like color drifts occur a 100%-monitoring of the colors would be helpful. To achieve this reliable monitoring devices which are supported by model based sensors should be developed. Model based sensors evaluate analytical process characteristics. With these process data and by means of system-embedded models it is possible to detect measurement errors, to adjust for them and to ultimately provide corrected values.
Color measurement technologies are using spectral sensors which show a UV/Vis spectrum with a particular resolution. This resolution, however, is quite coarse because there are only 40-70 measuring points covering the color spectrum. Color cameras provide more information on the color. They are, for example, able to separate the metallic flags from the colors since these influence the real tonal value and constitute a significant disturbance.
To determine the color, different areas are measured under standard light and from different angles: 15°, 25°, 45°, 75°, 110°. The determination of exact measurement points assures that different production parts are tested under the same circumstances and that consequently the user receives comparable data. Furthermore, factors like application conditions (e.g. angle displacement) or production parts with different color characteristics although the same paint (e.g. layer thickness) are considered.
For the modeling of the softsensors, first, the data from the data base (e.g. viscosity of the varnish, layer thickness) and the system parameters (e.g. angle, color value, temperature) are imported. Afterwards the data of the sensors are imported into the model so that it can calculate the real values in consideration of disturbing influences. Now the model is able to output real values, tendencies and predictions. Consequently, error sources can be recognized and localized.
By means of the error analysis, with regard to the modelized values, further possibilities of process control are available: sensor values outside the valid measurement range immediately indicate a process error. By fault isolation the erronous component can be localized. The fault type and its effects are detected so that countermeasures can be initiated.
By the use of this technology the delivery of imperfect production parts can be minimized. Additionally it is time saving and the user receives neutral and reproducible results.
In addition to providing the sensor data analytics and measurement value correction, divis can help you choose suitable sensors. As an independent company we are not tied to a specific sensor supplier. The decision is taken in close cooperation with the user. divis is always looking for the best cost-benefit-ratio.