Approximately one-third of the time taken to produce a vehicle is spent in the paint shop. Traditional manual and automated processes of paint application involve large numbers of steps with the final paint finish undergoing both automated and manual inspections. However, even with thorough inspections, it is estimated that 40% of the cars that exit a paint shop will undergo some form of rework of the paint once it has left the factory and before it is delivered to the customer. Research suggests that global automotive Original Equipment Manufacturers (OEMs) spend more than £75,000 on rework for every 10,000 vehicles produced. Hence, to reduce rework and associated costs and improve production efficiency, automotive OEMs are beginning to investigate and realise the benefits of predictive analytics and maintenance for paint application using the Internet of Things (IoT), Big Data Analytics, Predictive Analytics, and Artificial Intelligence (AI).

Benefits of Predictive Analytics and Maintenance for Vehicle Paint Application

Predictive analytics and maintenance within the automotive paint division are essential considering the value they add to the industry from a time and cost perspective. The main benefits of these are:

• Application failures and defects can be accurately predicted.
• Asset condition is continuously monitored with potential causes of defects identified.
• Continuous data collection is achieved.
• AI and complex algorithms are used to determine weak points in the system.

Predictive Analytics and Maintenance technology developments for Automotive Paint Shops

Dürr Systems AG, Germany, created Advanced Analytics in early 2020, the first market-ready AI application for automotive paint shops adopting predictive analytic techniques. This intelligent system identifies the source of faults and creates appropriate maintenance plans by combining cutting-edge Information Technology (IT) and mechanical engineering expertise. It also tracks previously unknown relationships and uses the information to tailor the algorithm to the plant using self-learning principles. Advanced Analytics is the most recent module in the DXQanalyze product line. The first practical deployment of this system suggested that the Dürr software improves plant productivity and the surface quality of painted bodies.

BMW, Germany, undertook a predictive analytics trial experiment in a paint shop at a BMW plant in Munich, southern Germany in late 2017 to identify paint problems using AI technology. As a paint quality assurance application, AI was chosen over traditional IT methodologies due to its greater ability for coping with both the volume and complexity of the data involved. In contrast to traditional IT, AI can extract cause-and-effect links from large and complicated datasets. The ability to analyse and use previously obtained data for current scenarios is the essential beneficial performance feature of AI. This trait distinguishes it technologically and economically from normal programming or human monitoring. BMW suggested that data from sensors and surface inspections enabled over 160 features relating to the car body to be monitored in real-time and that the quality of paint application can be predicted very accurately.

In early 2019, Taikisha Ltd, Japan, developed the i-Navistar system that was designed to analyse the root cause of operating failures and paint quality defects using IoT and AI technologies based on sensing data. By deploying this system, customers could optimise various paint production conditions and efficiently stabilise paint quality by accounting for the conditions of the entire production line. This led to a dramatic improvement in productivity and mitigated issues associated with the shortage of skilled workers in tight-tolerance production environments. The system works by utilising a real-time predictive analytics platform capable of collecting and structuring numerous types of data to perform data aggregation and paint anomaly identification. It also uses machine learning to automatically determine the status of system operations based on huge quantities of sensor data to execute extremely effective predictive data analytics and anomaly detection. Using novel techniques, i-Navistar provides realistic answers to previously difficult-to-solve problems, such as recognising failures and flaws that were hard to detect with standard threshold-based control systems.

The cutting-edge Honda manufacturing factory in Alabama, USA, employs the Splunk platform for predictive maintenance that consists of IoT, machine learning, and predictive analytics. It uses this system in all parts of its factory including the paint shop where it is used to forecast paint defects and loss in production efficiency, allowing Honda to produce over 350,000 vehicles every year. Splunk is a data platform designed for broad data access, advanced analytics, and automation. Unknowns and incidents associated with paint quality were essentially non-existent following this deployment at the Honda manufacturing facility. Furthermore, Honda has cut energy consumption and allowed personnel to focus on higher-level efforts through smart predictive maintenance using Big Data Analytics and machine learning.

ISRA VISION’s solutions for fully automated surface inspection in automotive production consistently and objectively detects and evaluates defects on almost all glossy automotive components. Painted surfaces are inspected in a cost-effective manner and defects can be located according to reproducible specifications. The CarPaintVision (CPV) surface inspection system detects and classifies all relevant paint defects on car body surfaces in a cycle-time-neutral manner, either in real-time or during production line stoppages. With modern hybrid predictive analytic technology, almost all car paint defects are detected at a capture rate of over 98.5%. Proven algorithms are used to inspect the data which is then classified according to the specifications outlined by each customer. Defect information is immediately available for continuous optimisation of the production process. However, it is possible to specify that any defects identified can be automatically marked for rework later.

Within the Audi Group, Germany, paint application data is collected from 2,500 sensors during the vehicle build. Here, valuable insights into this data are solved by storing sensor data in the data lake and processing the data with Apache Spark and Scala on an HDFS cluster. A plethora of random forest models are trained daily with MLlib to identify the most important drivers for paint quality for each failure and each layer of paint. The results are stored on an HDFS cluster and visualised with Tableau. To achieve business benefits, Spark is used alongside the whole process chain for data ingestion, transformation, and training in a productive and completely automated environment.

What Next?

The global predictive vehicle technology market is expected to increase at a Compound Annual Growth Rate (CAGR) of 20.1% from an estimated US$20.8 billion in 2019 to US$90.2 billion by 2027.

The following are the most prevalent obstacles that organisations face while implementing predictive analytics and maintenance technology: 1) The requirement for enhanced sensors, smart equipment, and advanced business analytics tools, 2) Enabling seamless communication across the many components of a Product data Management (PdM) solution, 3) System integration of IoT security, and 4) The challenge of high upfront expenses. Although these issues must be carefully examined when choosing a predictive analytics solution for paint quality, the benefits the solution provides, in the long run, will make it well worth the investment.

Finally, predictive analytics and maintenance will be as important to automotive OEMs in the future as enterprise resource planning (ERP) or financial planning software, since it allows for a level of equipment performance commensurate with demonstrating best practices, adhering to industry standards, and generating competitive advantages.