Skip to Main Content

Strategic Business Insights (SBI) logo

Internet of Things July 2019 Viewpoints

Technology Analyst: Guy Garrud

Predictive Maintenance in Semiconductor Manufacturing

Why is this topic significant?

Predictive maintenance is one of the most disruptive aspects of the Internet of Things. For complex manufacturing industries, it could prove to be invaluable.

Description

In June 2019, market-research firm MarketsandMarkets estimated that the predictive-maintenance market will grow from about $3 billion in 2019 to more than $10 billion in 2024. Companies with complex manufacturing operations stand to gain the most from predictive maintenance's advantages of optimizing downtime and minimizing damage from component failure. The semiconductor-manufacturing industry is in a particularly potent position to realize gains from predictive maintenance.

Memory-chip manufacturer Micron uses thousands of microphones and other sensors in its fabrication facilities to enable the company to detect and schedule maintenance needs. The company has also developed an AI system that uses convolutional neural nets to analyze the sounds that the thousands of microphones detect to identify what sounds emanate from which components and to recognize changes in these sounds that could indicate a component is close to failing.

Micron is also using sensors and AI to automate quality control and inspection tasks. The company uses a process it calls "real-time defect analysis," which is in essence a computer-vision algorithm for examining silicon wafers (a task that normally humans carry out visually). Overall, Micron estimates that it now has 25% fewer "quality events" and a 10% increase in production output resulting from its AI/predictive maintenance program.

Other companies are looking at offering predictive maintenance as a service. For example, Falkonry offers predictive-maintenance software as a service, providing machine-learning tools for industrial companies, including companies in the semiconductor industry.

Implications

One of the key advantages of predictive-maintenance systems over man-in-the-loop approaches is that AI and sensors are far less prone to fatigue and distraction than are humans when given a repetitive task. Indeed, machine learning can potentially outperform humans in identifying and monitoring data from machinery because it can incorporate input from multiple and diverse sensors. That said, analyzing large data sets is computationally intensive, and companies may face a trade-off between the granularity with which they can examine their operations and the computing power necessary to analyze these large data sets effectively.

One of the more interesting aspects of predictive maintenance is that for both machinery manufacturers and machinery operators, the associated data are potentially highly sensitive. Predictive-maintenance data could reveal the types of work a particular machine is tasked with or even show changes in manufacturing patterns over time. By the same token, careful analysis of predictive-maintenance data by machine operators could reveal information concerning how a particular component or piece of machinery operates. This issue is relatively minor if all of the maintenance work is carried out in-house, but in many cases, a machine's manufacturer will be better equipped to respond to maintenance issues than a machine operator is. Equipment manufacturers will be particularly keen to obtain predictive-maintenance data to train their machine-learning algorithms and to help design better and more reliable future products.

Impacts/Disruptions

Predictive maintenance is one of the most disruptive aspects of the Internet of Things but presents a potentially high barrier to entry. Companies with the data-analytics capacity to offer predictive maintenance as a service are likely to flourish during the next several years.

Scale of Impact

  • Low
  • Medium
  • High
The scale of impact for this topic is: High

Time of Impact

  • Now
  • 5 Years
  • 10 Years
  • 15 Years
The time of impact for this topic is: Now

Opportunities in the following industry areas:

Manufacturing, robotics, logistics, oil and gas, electronics

Relevant to the following Explorer Technology Areas:

Private IoT

By David Strachan-Olson
Strachan-Olson is a consultant with Strategic Business Insights.

Why is this topic significant?

Growing concern about user privacy is leading some technology companies to focus on providing advanced IoT capabilities with a focus on data protection and privacy.

Description

In recent years, the quality and quantity of smart and connected devices for the home have increased significantly. Most new televisions have smart features, many households now have smart speakers with virtual assistants, and other connected devices—including security cameras, door locks, and thermostats—are becoming more common. Although these devices offer connected features, they also introduce new privacy concerns for owners. Many smart-TV manufacturers rely on postpurchase monetization through advertising and data collection. Data collection includes the types of shows owners watch, the advertisements owners watch, and the owners' approximate location. Most smart speakers send recordings of their owner's commands to the cloud for processing and for indefinite storage. Many security cameras with smart-recognition features operate in a similar way. The cameras send video to cloud servers that analyze the images to detect people and objects, but companies often retain these data indefinitely. Multiple news outlets have published reports that employees and third-party contractors have obtained access to audio and video recordings as part of improving artificial-intelligence capabilities.

In contrast to most other major technology companies, Apple reports that it is making consistent efforts to develop technology that respects users' privacy. Apple's HomeKit IoT platform is privacy focused and continues to add new features while maintaining high levels of privacy and data protection. Apple recently announced HomeKit Secure Video, which will allow certain security cameras to perform smart recognition more securely. With HomeKit Secure Video, an Apple device in the home—such as an iPad, AppleTV, or HomePod—performs the image processing locally, analyzing the video from cameras to identify people and events, then encrypts the video and sends it to iCloud for storage. This means that Apple and third parties cannot see the video but that users still have access to image-recognition features. Apple is also working with some TV manufacturers to incorporate an Apple TV app that will not allow the television to access information about what a user is watching while using Apple's software. Privacy-focused IoT platforms and hubs are also available from smaller manufacturers such as Hubitat and Mixtile.

Implications

Deciding how to approach device and user privacy isn't easy for companies. Access to user data can provide companies with valuable ongoing revenue streams. Additionally, companies can use data to understand how people use their products and to train machine-learning models that enable new features. Emerging hardware and software technologies will make implementing AI features on device easier for companies, but companies may still choose to analyze and store user data.

Impacts/Disruptions

In recent years, individual privacy has become a growing concern for governments and users. After the introduction of the European Union's General Data Protection Regulation, numerous governments around the world began to consider similar data-privacy and data-protection regulations. In addition, many users have become increasingly aware of potential privacy issues since Facebook's scandal with Cambridge Analytica. Apple has begun to position itself as a privacy-centric company, but analysts are uncertain if this strategy will improve device sales. Potentially, increased demand for user privacy by individuals and governments could force companies to begin offering more privacy-focused IoT devices and solutions as standard.

Scale of Impact

  • Low
  • Medium
  • High
The scale of impact for this topic is: High

Time of Impact

  • Now
  • 5 Years
  • 10 Years
  • 15 Years
The time of impact for this topic is: Now to 5 Years

Opportunities in the following industry areas:

Home IoT, digital assistants, AI processors, cybersecurity

Relevant to the following Explorer Technology Areas: