This master's thesis covers the design, theoretical background and realization of a measuring system for the control of dimensional deviations of a car headlamp housing in a production line. Other than the car lens, the housing is the most important part in the tolerance chain (composition of independent dimensions and geometrical properties of deviations from nominal dimensions, i.e. tolerances, that are interconnected in a closed chain, sequence) of the cars headlamp. Since the dimensions of the housing can change during the installation of the internal components and during the clamping on the mounting stand, we designed a measuring system, that allows you to capture measurements directly in the process of joining the lens. The system uses a non-contact measuring method and is implemented on an existing headlamp mounting stand and consists of laser distance sensors, necessary peripherals and program code for capturing and sending data. The data obtained with the measuring system is stored in a database for further analysis. For the purpose of the analysis, two different algorithms for predicting the measuring points of the headlamp lens were created. The first algorithm is based on the specific derivation of rigid geometric mappings, and the second on the use of artificial neural networks. The algorithms are important because with the help of measuring points on the lens we can control the geometric accuracy of the headlamp.
In the introductory part, we describe the theory of tolerances and process control. This background is necessary for understanding and planning the system for such control. We also describe the given coordinate systems and correlations between them for the needs of understanding and further data processing. We also delve into our examples tolerance chain and dimension control of the headlamp, where we describe the individual relevant steps in the headlamp manufacturing process.
The central or empirical part is focused on a specific project and headlamp. We describe the evaluation, selection and implementation of sensors for our measuring system. After this part, the capturing and processing of data obtained by sensors is explained including all calculation methods, software and hardware. We also present an Excel based database created with the Python language, which is intended for storing and display of results. Theories for both methods of predicting lens measurement points, one using rigid geometric mappings and the other with neural networks, are also described.
In the conclusion we descriptively and graphically present data obtained with the developed system. We analyze a fast capturing of the process of gluing the lens to the headlamp and single captured "online" data over a longer period (several months). In this paragraph, the importance of data storage and analysis for fault detection in the production process are shown on a real-world example. We also describe the testing and analysis of both methods for predicting the measurement points of the lens, compare them and validate them with the already created reference measurement method in the company. Finally, by analyzing the obtained data of the created system and the chosen method, we create an improvement on the headlamp housing mounting stand and improve the dimensional stability of the headlamp. We also describe possible improvements and the benefit of using the presented measurement system for the company.
|
