AIoT Drives Automotive Camera Production to Excellence

What is AIoT?

The Fourth Industrial Revolution (Industry 4.0) is essentially the Digital Age, characterized by a heavy focus on automation, real-time data, connectivity, embedded sensors, and machine learning. Its iconic representation is probably the smart phone – a powerful handheld computer extending the power, reach and versatility of the Internet to all corners of the globe. In tandem with the smart phone’s revolution of consumer and business life, a similar wave of change is underway at the factory and production floor, offering the promise of radically transforming the industrial space with a vision for the Smart Factory of the future.

However, definitions of a Smart Factory vary - forcing manufacturing companies all over the world to grapple with uncertainty about the dramatic changes they need to plan for to help their business thrive in the future. With more demanding quality, safety and reliability criteria to fulfil, the irony is that despite better machine inter-connectivity, real-time data acquisition and inter-factory communications, most of the time, data is still being stored in too many unintegrated places – in machines, at customer sites, in the cloud, and so on. This has made reconciling data a challenge, and generating actionable insight from such data even more difficult. When we add the exponential increase in data generated in a more connected world, the existing data management paradigms become unsustainable.

At the same time, the confluence of these computing capabilities have also made the evolution of the next generation of industrialization a la ‘Industrial 4.0’ possible, from which the ‘Internet of Things’ (IoT) is derived from. The IoT represents a vision of a connected world of devices talking to one another: coordinating, cooperating and scheduling tasks.

However, the future is beyond just information collection and analysis, it must incorporate machine learning ability. The marriage of IoT and Artificial Intelligence (AI) into an ‘AIoT’ ( Artificial Intelligence of Things) involves infusing AI capabilities into ‘smart assembly lines’ that enable equipment to independently examine data, perform analysis make and act on those decisions. The architectural vision of a full automation production environment with efficient supply chain management and optimization, and predictive maintenance of factories for increased efficiency and volume output, is a compelling one - delivering data-driven closed loop insights and automation without the need for human intervention.

How AIoT Augments Automotive Camera Production

This drive towards zero PPB (Defective Parts Per Billion) and near-zero unscheduled downtime (or higher Overall Equipment Effectiveness (OEE)) for manufacturing environment is being driven by a host of factors including avoiding high product recall & repair costs, higher safety & security requirements, and increased competitive pressures. According to Deloitte, product recall, repair costs and unplanned downtime can cause industrial manufacturers to lose about US$50 billion per year (not to mention effects on company reputation and customer experience).

Figure 1: Image via Shutterstock

The automotive industry holds safety and quality as key product criteria, because their products affect the lives and safety of users and more. Years before, the automotive industry introduced ADAS (Advanced Driver Assistance System) in order to raise safety capabilities to reduce loss of life and significantly minimize financial losses due to recall of unsafe or defective automotive products. ADAS technology revolves mainly around vision-based (camera and LiDAR), frequency-based (RADAR) and acoustic-base (SONAR) sensor technologies to accurately detect the surrounding environment for a variety of purposes.

Within the context of vision-based camera technology, the build quality of the automotive camera unit is critical to the clarity of the surroundings visualized by the vehicle. Over the years, particular cases of automotive OEMs models recalls were due to safety issues related to the automotive cameras. Besides the safety issues, these recalls resulted in millions, if not billions of lost dollars in damages.

Some recent incidents:

• 2019-2020, Mercedes-Benz: 375,843 vehicles recalled due to defective rear view cameras1
• 2020-2021, Ford Lincoln: 34,975 vehicles recalled due to defective rear view cameras2

AIoT technology can help augment automotive camera production factories in many areas encompassing performance, inputs quality, resource allocation, process consistency and refinement, among others. Here are some examples:

• Managing and analysing the input quality and consistency of incoming materials, for example, image sensor PCB, camera lens quality and more.
• Analytics compensation of process parameters control and feedback from downstream processes, for example, MTF variation post-alignment process feedback from end-of-line tester systems to compensate alignment offset parameters.
• Reducing dependence on human operators in process determination, via data analysis of parameters of choice and duplication of ‘golden recipes’ to corresponding machines.
• Providing predictive maintenance of machines via diagnostics and analysis of machine performance and trend predictions.

Figure 2: Image by ThisisEngineering RAEng via Unsplash

All the above contributes greatly to the integration of the 4 Ms (Man, Machine, Materials, Method) in automotive camera production:

• Man – streamlining manpower required for the various parts of the production ecosystem, from machine operation, scheduled downtime for maintenance, process and parameter optimization, material usage monitoring, and supply replenishment.
• Machine - machine operation consistency from PCB assembly, optical module and image sensor alignment, to camera unit assembly and validation before delivery. Enabling closed governance on yield and throughput performance.
• Materials – optimising and ensuring the flow and quality of incoming materials into the system, which in turn affects the quantity of parts processed as final products. Reliable, continuous supply of materials is critical for the production floor to operate perpetually and efficiently.
• Method – the parameter and process criteria that equipment use to operate and validate the input and output results. In automotive camera production, this can encompass the selection of the required parameters for optical elements sensor alignment settings and process criteria for die placement repeatability (the ‘golden recipes’). This enables consistently high throughput and high yield output.

Figure 3: Image by ASMPT

Different Stages to Artificial intelligence Production Environment

These descriptions may seem to be a ‘Shangri-La’ level of attainment for production environments to undertake, but clear implementation steps can makes this very achievable.

Figure 4: Image by ASMPT

Reactive & Preventive Maintenance

Preventive maintenance commonly practiced by majority of automotive camera production houses, has typically involved scheduled downtime for maintenance and checks of these equipment. Despite this, unscheduled downtime can still happen when unforeseen circumstances occur, for example, for machine recovery, material shortage, material quality evaluation or process optimization. These delays inevitably put the entire system’s manufacturing efficiency and output fulfilment abilities to the test.

In addition to properly-planned scheduled downtime for maintenance and evaluation of an individual machine efficiency, some additional steps that can be taken for improved management include:

• Digitalization of machine data and connectivity to a central system for overview management
• Incorporating KPI management tools for criteria control and implementation of a Recipe and Process Control management setup
• Remote Monitoring Control for operator off-site remote access to machines

At this stage of control and management, the majority of tasks still utilise manually-based evaluation and decision making based on fairly ad-hoc, reactive considerations of the production environment.

Conditional Based Maintenance

We can augment this by placing the maintenance criteria for machines and equipment under statistical or rule-based monitoring capabilities. Production efficiency and continuity can be enhanced using such diagnostic elements, basically by allowing the production controller to evaluate all 4Ms key elements and their effect on yield and OEE performance.

Conditional-based maintenance capabilities are another area. They can utilise data analytics capabilities for a production situation, with a holistic evaluation of the events that occur with deep analysis of the data in order to generate nuanced and more targeted insights for decision and action. Consider:

• Machine health management for real time data processing of production machine health and performance, to determine the next action to take.
• Anomaly evaluation via closed loop feedback of testing to assembly for assessment of output disparity and analytics-based assessment of process parameter optimization.
• A Data Analytics overview of production performance especially the 4Ms elements, evaluating through data input of all factors to detect the root cause of performance reductions.

With the ability to acquire statistics and data of the 4Ms factors on hand in a production environment, decision making via analytics of production process and outcome can be made with improved consideration and accuracy of action to be taken. Improved utilization of machines and closed loop feedback on golden recipes optimization can lead to better control of production yield and quality

Predictive Maintenance

This refers to the provision of true predictive maintenance capabilities via machine learning analysis of big data acquisition from prior stages of connectivity implementation. AI through machine learning and data analytics allows manufacturers to predict production issues before they actually happen, enabling timely pre-emptive corrective actions to be taken to ensure production continuity and high quality output. AIoT in turn enables full automation of the entire production analytics, leading to automated preventive maintenance capabilities effectively streamlined via data and rule based evaluation; this in turn allows predictive capabilities to improve overall production outcomes proactively.

Consider:

• Smart Factory enablement allows automotive camera production automation and connectivity attaining Industry 4.0 standards
• AI through machine learning and big data analysis gives rise to predictive decision and optimization improvement to process parameters optimization, material resources allocation, machine performance sustainability and maintenance efficiency, and manpower reduction and human error minimization.
• With cloud or HMTL accessibility, satellite remote access and control for round-the-clock insight, removing geographical constraints on production status and analysis.
• Optimizing risk management on machine malfunctions, material quality inconsistency, process parameters variations, manpower operation deficiency and more.

ASMPT ACamLine and SkyEye enables the Automotive Camera Production Path to Excellence

As part of ASMPT’s latest development in the automotive camera assembly line, ACamLine embodies the capability of full camera assembly process for high yield and high volume throughput. Its modular platform design approach provides the flexibility in machine configuration to suit every customers’ production sequence and ease of scalability for ease of investment of customer in different stages of their production volume needs.

AQUA

Figure 5: ASMPT ACamLine, Image by ASMPT

The core of the AIoT solution is powered by the Group’s SkyEye software engine, which uses advanced machine learning algorithms to process manufacturing data. This platform enables customers to start their AIoT journey via several progressive entry points. From first getting factory tools connected onto an AIoT ecosystem in order to improve performance autonomously; enhancing current rule, and event-based activities with AI to enable predictive maintenance and procedure tweaks in order to improve OEE and yields. Finally, an advanced path through the adoption of AI and Data Analytics tools enable predictive capabilities that can prevent production process fallout. Crucially, customers will be able to scale these capabilities on their own terms both locally and globally via an ASMPT Cloud Service to manage multi-site environments. ASMPT SkyEye - the newest development of multi-tier cloud based production and equipment management system embodies the true essence of integration and synchronization of 4Ms factors. Along with empowerment of machine learning and machine data acquisition to achieve powerful Artificial Intelligence production management and control.