Data-to-value solutions are best navigated with a cross-functional approach integrating different domain expertise. This premium automotive OEM collaborated around data-driven optimization to achieve digital era continuous improvement robotic stud welding.
AUTOMOTIVE MANUFACTURING IN THE DIGITAL AGE — TURBULENCE AND INNOVATION
Quality welding in automotive manufacturing is part of cost-efficient production. Efficiency gains in discrete processes are welcome relief from broader industry challenges, including:
- a semiconductor shortfall
- evolving supply chains
- expanding environmental rules
- escalating energy costs.
Nevertheless, there is a critical constant for an automotive industry under stiff market headwinds. Provide high-quality components and parts more efficiently. Yet, fulfilling the promises of digital era production means mastering complexity.
QUALITY WELDING PERFORMANCE GAINS WITH PRESCRIPTIVE AI
Powered by the historical production data manufacturers already own, AI insights can fast-track value. Critically, digitally enabled manufacturing process optimization is attainable with minimal capital investment.
Consider the example of this premium vehicle manufacturer. They optimized the robotic stud quality welding performance by utilizing the dynamic guidance capabilities of prescriptive analytics. Moreover, the manufacturer accomplished this in a short time frame. Measurably enhanced quality outcomes, reduced defects, and better uptime were among the value-adds. Here are the key results:
- A reduction in downtime of 72% within the first month.
- Increased production volume.
- Estimated monthly cost savings of between $120,000 and $140,000.
THE NEED TO BOOST AUTOMOTIVE QUALITY WELDING PERFORMANCE WITH DATA-DRIVEN TECHNOLOGY
Automated spot welding is inherently complex. It involves numerous variables, often leading to welding faults. These faults result in kick outs, production delays, and increased rework costs, significantly impacting operational efficiency and welding quality.
In addition, traditional statistical solutions face many complexities in optimizing robotic stud welding output. Consequently, accounting for myriad variables circumscribing automotive welding outcomes can be daunting for plant engineers.
Blindspots in the data present a practical challenge for OEMs. Identifying the root cause of anomalies to realize and sustain new quality benchmarks with traditional analysis stretches capacity.
Let’s put this challenge in an automotive quality welding context.
A ROBOTIC STUD QUALITY WELDING CHALLENGE
The body shop is one of the most complex areas in this automotive manufacturing process, with at least 200 robots operating. Ultimately, the main barrier to peak operational efficiency for this manufacturer was significant downtime. The downtime was due to stud welding faults, resulting in substantial costs to the business.
Fortunately, data scientists with manufacturing process experience can work cross-functionally, collaborating in automotive plants to bridge the skills gap. Ideally, functional manufacturing experts and industrial technology vendors work together on a clearly defined performance optimization challenge.
A COLLABORATIVE APPROACH TO MANUFACTURING PERFORMANCE OPTIMIZATION
In partnership, manufacturers and AI specialists can turn complexity into a clear and actionable picture of an automotive manufacturing process. In this case, DataProphet sent a dedicated team of data science experts to the automotive stud welding facility.
An initial data collection phase used several months of historical production data mapped to known quality outcomes. Data scientist and mechatronic engineer Jan Combrink explains the importance of comprehensive traceability and continuity in the data collection phase:
“Essentially, the view you want is a complete mapping of the process from input A to result Z. The last variable in the chain is the quality outcome; quality is traced back to the last sub-process. And from then on, that last sub-process has to be traced back to the previous sub-process — back to the system’s input.”
From here, an AI-enabled process performance optimization solution can be deployed. However, because no two manufacturing facilities are alike, DataProphet also considers the uniqueness of the client’s production facility. Seamless integration with existing data and IT infrastructure ensures a fit between the IIOT platform and production dynamics.
Figure 2 assumes this integration is in place and provides a high-level overview of the critical interdependencies of the robotic stud welding process in this use case. It highlights the key targeted parameters leveraged for data-driven optimization.
Let’s take a moment to consider the core technology at work here.
PRESCRIPTIVE AI — THE RIGHT ENGINE FOR AN AUTOMOTIVE PROCESS PERFORMANCE EDGE
When prescriptive AI connects an underlying control plan with a desired production outcome, KPIs can be targeted for improvement. These might include quality, throughput, or overall equipment effectiveness. Machine learning engineer, Luyolo Magangane, explains deep learning’s superiority for production optimization:
“It comes down to a mathematical property. Artificial neural networks enable the most efficient representation of any arbitrary mathematical function or probability distribution. The capacity to explain all interactions and interdependencies is not true for any other type of machine learning algorithm. Deep learning discovery is particularly well-suited for complex processes such as manufacturing.”
OPTIMAL PRESCRIPTIONS FOR THE STUD WELDING QUALITY
Complex interactions underpinned our client’s stud welding process. Many process variables influencing the quality of their welds were considered. They included:
- weld time
- amperage
- lift height
- stick out
- plunge speed.
DataProphet PRESCRIBE also identified over 800 stud types used in the facility’s welding operations. Therefore, describing optimal process parameter values for all of them had to factor in both the stud type and these variable interactions. With a comprehensive production data picture mapped to quality output, the complex interactions were explainable via prescriptive AI modeling.
The automotive manufacturer then produced a dynamic control plan with the DataProphet team. This plan included the following:
- 1. Adaptive control limits for each welding parameter.
- 2. Recommendations for two car models’ optimal set points.
As a result, the plant engineers and operators were confident the recommendations could improve weld quality. Applying the prescriptions substantially reduced downtime imposed by defective welds.
As mentioned, these improvements translated to monthly cost savings of between $120,000 and $140,000, saving the automotive manufacturer approximately $4.5 million dollars per year. Welding faults dropped by 55% over the course of a month, and weekly Q-stops went from 81 to 20 (a 72% reduction in spot welding related downtime).
Ultimately, the prescriptive AI solution helped the client maintain superior quality in a facility that already enjoyed high-quality maturity. Better yet, the manufacturer achieved this purely by leveraging its own data.
THE ROAD AHEAD FOR DIGITAL ERA AUTOMOTIVE PRODUCTION
Data-driven technology can be easily scaled for AI-driven performance optimization to accommodate additional automotive environments for which some historical production is available. Rolling the solution out ensures consistent optimization and quality improvement across the factory group.
As the automotive industry evolves, data-driven solutions will become the norm for optimizing manufacturing performance and enhancing product quality. By leveraging prescriptive AI for other products and at other plants, automotive OEMs can gain an early-mover advantage.
A deeper understanding of the complex interdependencies within processes is an added benefit. Strategically, manufacturers embed the normalization of data-driven decision-making. This smart factory step change primes automotive manufacturing teams for the digital era.