Optimising the painting process: Data Science and Big Data in the Paint Shop

from | 24 September 2019 | Basics

Optimising the painting process and avoiding errors in the long term - these two goals can be achieved with Data Science and Big Data in the paint shop. In this blog article, we offer an insight into one of our projects. In the process, we explain basic Requirements in data science projects and show the potential of data science for the automotive industry.

In the production and Assembly a Car bodywork there are numerous error-prone processes that are time- and cost-intensive are. The painting process is an essential and at the same time error-prone component. error-prone component. A car body often passes through the painting process several times. process several times. Individual parts even have to be touched up manually. This raises the question of how the Painting process optimise lets.

The IT-supported use of Big Data analytics respectively Data Science enables the entire Production process holistically and to optimise it sustainably. This makes it possible to eliminate sources of error and to recognise connections that were previously not obvious. In this way, a extensive optimisation potentialthat makes it possible:

  • systematic Error to avoid,
  • the Production in Real-time to review and
  • the Production process in the long term improve.

The challenges of data science projects: Understanding the underlying processes

At one of our Customers, an a car manufacturer, there were always difficulties in the bodywork difficulties. After the painting of the individual parts, the initial assembly of the body was and in many cases the parts did not fit properly or the gaps did not fit. or the gap dimensions did not fit. That is why it came to time-consuming, manual reworkingwhich subsequently had to be carried out.

The assumption was that the dimensions of the gaps and joints had been joints had changed so much due to the painting process or due to the paint being applied too thickly that that the nominal values were exceeded during installation. were exceeded.

The painting process at a car manufacturer (here using BMW as an example).

At the same time Defects in the paintwork one of the most common triggers that later cause damage to the bodywork. In order to optimise the painting process, it was first necessary to clarify a few basics. had to be clarified. The challenge in this case is to gain a more precise understanding of the understanding of the painting process. This is the only way to identify all the decisive for Error analysis and -avoidance as well as Possibilities for Automation and Process optimisation Identify.

Search for suitable sources for data collection

Another key challenge in data science data science projects is to develop a solid Data basis to create. At the beginning of every analysis project there is therefore the crucial question of the Data sources. Particularly in the case of processes that are not yet processes that are not yet digitised or that are difficult or impossible to digitised, this question poses a particular challenge.

In the case of the car manufacturer's paint shop, exactly this case. The paint is applied there with a Painting robot in three phases. In order to obtain a data basis of this process, the entire process had to be broken down into small components.

The individual components are then assessed for their relevance and their suitability as Data collection points examined. To obtain these points have to be partly equipped with appropriate Sensors be equipped. This results in a whole range of Parameters, which are in the Analysis process be included: The amount of paint the amount of paint used per coating process, the PH value of the paint, the pressure during the drying temperature or data on the dwell time at certain stations. stations. This principle of data collection can be transferred to various other applications. applications. The question is always which data sources are available and which and which additional data sources can be tapped.

Link tip: The Data provision is one of the central components of data projects. Read all about it in our article on Data pipelines.

Data modelling, big data and holistic data processes

The first step on the way to a solution is the Creation A modelwhich should describe reality as accurately as possible. A technical data model is created as part of a data science process to is created in order to optimise the painting process, as in this case. To do this a data basis must first be created and then constantly updated. updated. This means: Big Data analyses are iterative processes.

In some cases, they make it necessary to collect data on a large to permanently collect data on a large scale in order to obtain as accurate a picture as possible of the entire process. of the entire process. An analogy to illustrate this: based on a single photo of a car, it is not possible to say whether it is a single photo of a car, it is impossible to tell whether it is moving, stationary, accelerating or braking. or braking. However, if the Data basis to 24 frames per second, we get an exact image of the car's behaviour.

The larger and more precise the data basis, the more exact the the analyses and Forecasts via current and future behaviour in reality. In the case of production of the car body, data collection and analysis takes place on three levels:

  • Firstly through the Measurement of the lacquer film thickness
  • Secondly, through the Measurement of the gap and joint dimension
  • Thirdly, through a Paint Defect Analysis on the basis of optical image analyses

Through its holistic approach, this Data model to Findingsthat were not expected. For example, a direct correlation between a single parameter of the coating process and the and the change in the gap and joint dimensions.

The solution: visualisation, forecasting and control in real time

In the solution for the car manufacturer, the a Variety of sensors Targeted at the relevant points Data raised and thus the entire Production process documented. Building on this database, a data model could be developed and a comprehensive technical and professional comprehensive technical and professional understanding of the data. With With the help of the model, it was possible to identify the significant Influencing parameters for Prediction of the optimum coating thickness Identify.

This is now visualised in real time so that Connections and Anomalies defects and coating thickness can be identified immediately. identified immediately. Through the Real-time visualisation can to determine at any time when and how the coating process has to be adjusted to obtain the correct gap and joint dimensions.

The big advantage: Already during the process of quality control can be carried out during the be carried out. Errors are detected earlier, causes can be eliminated immediately and the quality of the result is and the quality of the result can be sustainably and significantly increased.

The result: Fewer errors, a higher degree of automation and a better understanding of the overall process

In the end, various Process parameters which have a significant influence on the coating thickness and optimise the coating process. and optimise the coating process. The improved understanding of the process as a whole, causal relationships could be identified. The Data analysis and -modelling led to the necessity of permanently collecting data process to be subjected to permanent data collection.

At the same time, other sub-processes, such as the the Quality control of the painting process to automate the process. Measured and target values are already compared during the painting process. with each other. Through the analysis on granular level could even visualised further irregularities in the individual layers of paint. could be visualised.

Link tip: Read our article here about Advantages and opportunities of digitalisation for SMEs.

Optimise the painting process and sustainably increase quality

Due to these measures, it was possible to optimise the painting process. optimised. Thanks to the wide range of defect measuring stations, several different several different defect patterns could be determined. The Quality of the lacquer could be significantly increased as a result. In the the result, it was possible to determine the defect intensity depending on the and to reduce it in part automatically.

This is achieved in particular through the availability of real-time predictive action recommendations. predictive recommendations for action are available. The Algorithm determines the optimum setpoints for car body construction and creates a forecast for the paint thickness. The rework that was often necessary in the past often necessary to date have been reduced to a minimum, while at the same time the quality of the paint was increased at the same time.

Would you like to learn more about data science? Sign up for our Data Science training.


Michaela Tiedemann

Michaela Tiedemann has been part of the Alexander Thamm GmbH team since the early start-up days. She has actively shaped the development from a fast-moving, spontaneous start-up to a successful company. With the founding of her own family, a whole new chapter began for Michaela Tiedemann at the same time. Hanging up her job, however, was out of the question for the new mother. Instead, she developed a strategy to reconcile her job as Chief Marketing Officer with her role as a mother.

0 Kommentare