AI in Manufacturing: Diving Into Industry Problems and AI/ML Solutions
Table of Contents
AI technology is here to make some significant changes. However, just like every recent invention, it’s hard to navigate the amount of information and find truly productive solutions for the specific business. So, where to start your exploration? IBM surveyed 4514 enterprises across the US, EU, and China to check their progress in AI exploration and adoption:
AI Adoption Rates in Businesses:
- 34% of businesses in the US, EU, and China have deployed AI.
- 39% are in the process of ramping up exploratory phases with AI.
AI Implementation Strategies:
- 40% of respondents deploying AI are working on proof-of-concepts for AI-based projects.
- 40% are utilizing pre-built AI applications like chatbots and virtual agents.
AI Adoption by Company Size:
- 45% of large companies (1,000+ employees) have adopted AI.
- 29% of small and medium-sized businesses (under 1,000 employees) have adopted AI.
Barriers to AI Adoption:
- 37% of respondents cite limited AI expertise or knowledge as a hindrance to successful AI adoption.
- 31% face challenges due to increasing data complexities and siloed data.
- 26% lack tools for developing AI models.
Primary Uses of AI in Companies Currently Deploying:
- 36% prioritize data security using AI.
- 31% focus on the automation of processes through AI.
- 26% utilize virtual assistants and chatbots.
- 24% engage in business process optimization using AI.
- 24% perform sensor data analysis and utilize the Internet of Things (IoT) with AI.
What is AI in Manufacturing?
Artificial intelligence (AI) in manufacturing refers to applying advanced computer algorithms and technologies to improve and optimize manufacturing processes. It entails using machine learning, data analysis, and automation to make operations more efficient, accurate, and adaptable. AI tackles the challenges across most manufacturing stages, from product design to quality control.
Common challenges of the manufacturing industry
Among the hindrances on the way to AI adoption is a lack of expertise. Does that mean that an individual organization without a strong IT background and expert team cannot afford AI? If one industry fails to implement AI, does it mean it’s a failure for all? Not at all. AI application requires an individual approach to the organization regardless of the industry. Manufacturing is actually the industry that can benefit from AI the most since it generates approximately 1,812 petabytes (PB) of data annually, and AI thrives on data.
The issue of managing data in manufacturing has been looming for a long time. Ignoring it leads to:
- Quality control issues
- Poor maintenance
- Lack of continuous improvement
- Inability to adapt to market changes
- Lack of process optimization
- Inaccurate decision-making
Considering the challenges the manufacturing industry faces, business owners resort to new ways of solving them. Among them are the following technologies:
- Robotics and automation
- Data Analytics
- Internet of Things (IoT)
- Additive manufacturing
- Cloud computing
- Artificial intelligence/machine learning/cognitive computing
- Digital twin
- 5G connectivity
- Augmented/virtual reality
- High-performance computing
- Edge computing
- Quantum technology
What are Artificial Intelligence, Machine Learning, and Cognitive Computing?
Let’s start the journey of selecting the AI solution for manufacturing by defining the basic terms.
Artificial intelligence (AI) is a technology that allows computers to think like humans. It enables human-like reasoning and problem-solving capabilities and lets computers understand natural language, recognize patterns, make predictions, etc.
Machine Learning (ML) is a subset of AI, and it focuses on designing algorithms and techniques that allow computers to learn from data and improve their performance on specific tasks with time. So, instead of being programmed to perform a task, the ML system learns from practicing and can make predictions, and decisions, and solve problems. In essence, it is about teaching a computer to perform tasks by showing different examples and letting it learn independently.
Cognitive computing is another subset of AI that focuses on simulating human thought processes and cognitive capabilities. It creates systems that learn, reason, understand, and interact with humans naturally and intuitively. Cognitive computing systems use context-awareness natural language processing to understand and respond to human language and behavior.
Artificial intelligence is about making computers smart. The ultimate goal of AI creators is to make it as smart as a human, and machine learning is a technique used to train the computer on examples. AI and ML have a wide range of applications, while cognitive computing is a subset geared towards improving human-computer interaction.
How to use AI/ML solutions in manufacturing?
While the tech industry leaders are racing to win the AI competition and design the smartest computer, other businesses often get confused about their AI strategy. Where should a small business find AI/ML services? Does it mean every business now needs a strong in-house tech team? The easiest way for a manufacturing facility to get hold of AI is to use pre-trained solutions from one of the industry leaders like Google, Amazon, Microsoft Azure, or Open AI.
These companies offer accessible and powerful solutions that let manufacturers harness the benefits of artificial intelligence without the need to develop their models or invest significant resources into training. Check out the table below for a comprehensive list of companies, their AI products, and the specific AI in manufacturing examples:
What is the cost of disruption in the manufacturing process? The latest estimates show $50 billion annually. Moreover, the productive capacity of manufacturing can be reduced by 5 to 20 percent due to maintenance issues alone. Knowing the cost of this problem, manufacturers are looking for solutions that can improve the situation dramatically. However, the effectiveness of the known solutions is not game-changing. So far, this search for a remedy has led to preventive and time-based maintenance. Why are these solutions not helping?
Well, it puts you as a factory owner in a position where you have to take the machine off duty for service and cause downtime, or you can max out the capacity of each part and replace it when it breaks. How easy is it to replace parts? Traditionally, it means that you get yourself into ordering parts, waiting for their delivery, and potentially looking for professionals who can install them. Hence, unplanned downtime and expenses. Predictive maintenance is the alternative that comes with introducing technology into manufacturing. The Internet of Things in manufacturing is probably the best example of affordable and productive technology that has changed operations. In short, it connects physical assets to servers and translates physical actions into digital signals. Devices like temperature, vibration, and moisture sensors collect vast amounts of data that are transformed into analytics.
Using one technology might cover only some of the set of needs of the enterprise. The Internet of Things is a perfect solution for gathering vital data. The next step is to decide what processes you want to improve or what resources you need to focus on to implement the right technology to perform specific tasks and put the data into action. How is AI used in manufacturing, then?
Suppose you are at the stage of using some bits and pieces of technology. In that case, it’s possible to integrate AI/ML algorithms to accelerate improvement and increase the efficiency of a manufacturing facility. When it comes to predictive maintenance, AI/ML is applied to predict equipment failure based on historical data, as well as optimize maintenance schedules to reduce downtime. The most advantageous course of action is a seamless integration of AI/ML with the Internet of Things (IoT) infrastructure. This integration facilitates the continuous reception of real-time data streams and enables the application of sophisticated analytics, empowering organizations to derive actionable insights and make well-informed strategic choices.
A factory furnished with advanced technology facilitating communication between machines and between machines and humans, coupled with analytical and cognitive capabilities, empowers accurate and timely decision-making.
When implementing AI/ML-enabled predictive maintenance, Caterpillar, a construction and mining equipment manufacturer, created advanced analytical tools. With their help, they developed failure models. One of them is engine oil dilution. Previously, the incident detection took up to 10 days, and with the AI/ML tools, this time was reduced to 2.4 hours and saved approximately $ 360,000.
Another algorithm brought by AI/ML capabilities is object detection. It’s like using a smart eye to quickly detect defects like missing parts or scratches on the product. Instead of people performing the inspection, safety check, or quality control, the algorithm does it faster, cutting down the time spent on production.
As a part of its AI strategy, Ford Motor Company introduced Latitude AI to design self-driving cars. However, in addition to using AI capabilities in their product, they also enhanced the manufacturing process. Object detection was used to identify wrinkles in the car seats. It was one of the points of Ford’s quality assurance protocol performed by employees and automated with AI.
The PPE detection model is used to confirm the presence of protective equipment to minimize exposure to workplace hazards. Integrated into the cameras or sensors around the facility, the AI algorithm identifies PPE items like face masks, gloves, and helmets and checks if they are worn properly. As workers enter the workplace or facility, the system can detect the lack of PPE and send alerts to the supervisor, inviting them to intervene. Another benefit of such a system is the capability to provide real-time video monitoring. It means that the videos from cameras of data from sensors are analyzed continuously and not post-factum.
Honeywell ThermoRebellion system designed to screen workers’ temperature employs a PPE detection algorithm to ensure that personnel entering the facility wear protective masks. While the primary aim of the solution is to check temperature using cameras and advanced artificial intelligence, it also provides information about the presence or absence of protective equipment.
AI-enabled occupancy analysis allows you to count people within the desired areas. However, with the fresh AI capabilities of Google Vertex AI, for example, you can do even more than simply check the availability of one room as it detects:
- Line crossing. Since manufacturing often involves hazardous work conditions, business owners must follow Health and Safety Regulations. Using line crossing, you can prevent unauthorized access to certain areas.
- Dwell time detection. In addition to counting the number of people entering the area, the algorithm identifies human presence and activity. Tracking the number of people in different factory zones enables efficient resource management. For example, you can adjust lighting or temperature to save energy in the spaces not occupied by personnel during the shift.
Getting real-time data on occupancy does not require special sophisticated equipment as it can use standard security and surveillance cameras that most facilities already have.
Siemens Amberg Electronics Plant is an example of a futuristic automated manufacturing facility that uses occupancy sensors in combination with AI algorithms to collect and analyse historical data, and real-time patterns, and predict the occupancy in different areas of the plant. The algorithms use occupancy data to optimize production line operations. For example, it predicts the number of employees expected in each zone of the plant during shifts. These predictions are used to adjust the workforce allocation and production schedules.
With person/vehicle detectors, manufacturers can achieve greater operational efficiency and enhance safety. The algorithms enable precise identification of people and vehicles as well as their movement around the designated area.
AI-enabled person/vehicle detection is implemented into the existing infrastructure of the facility and is combined with cameras and sensors to provide real-time updates used for:
- Safety enhancements
- Equipment and asset tracking
- Regulatory compliance
- Workflow insights
- Detection of unauthorized access
- Intrusion detection
Person/vehicle detectors are popular among manufacturers that deploy Automated Guided Vehicles (AGV) and robotics. Amazon is testing AI/ML solutions at its fulfillment centers around the world. The uniqueness of Amazon warehouses and fulfillment centers is that robots or AGVs work alongside humans. To prevent accidents between machines and make sure that robots can navigate the facility freely, they use a Person/Vehicle detection system powered by Amazon Rekognition.
Product design is complicated due to the multifaceted challenges and considerations involved. On the way to bringing a concept from ideation to a tangible product, the creators face regulatory compliance, technical constraints, diverse stakeholders, and material selection. Hence, the designer has to accept the trade-offs to balance the desired functionality, aesthetics, cost, usability, and market appeal. AI/ML algorithms optimize the product design process, considering all the restraints and requirements, and suggest ideas much faster than a human can. Traditionally, engineers used multiple calculations simulations, then performed iterative testing and adjustments, leading to increased use of resources and time. Instead, the options generated by AI can be quickly tested and validated.
The subsidiary of BMW, Designworks, has experimented with AI capabilities in design. While not all the ideas were successfully implemented, wheel rims and car seats were created using generative design techniques.
AI-enabled material science entails the integration of AI into the field of materials research and development. It accelerates the process of discovery design and optimizes materials with specific properties and characteristics. Hence, manufacturers are one step closer to creating innovative, groundbreaking products and solving most of the material-related problems.
Toyota is an example of a company that harnesses all the AI capabilities, in particular, to develop lightweight and high-strength materials for their automobiles. To be specific, they used AI to analyze the properties and characteristics of the materials and integrated this information with the design requirements (weight reduction, safety, and durability). As a result, the vehicle’s performance objectives were aligned with the choice of materials.
During the COVID-19 period spanning from 2021 to 2022, a total of 123,000 workers in the UK reported contracting the virus in their workplaces. Meanwhile, 565,000 sustained injuries. The cost of such incidents reached 7.6 billion pounds. In addition to slips and trips, workers often deal with dangerous machinery and equipment. Hence, the need for strict safety protocols arises. Monitoring the machines is even more effective in incident prevention. Advanced algorithms continuously monitor the activities of workers and the environment in real time. In that way, hazards and unsafe practices can be detected and addressed promptly.
COVID-19 labor safety solutions have gained significant popularity. Honeywell, an American company that manufactures security, safety, and energy solutions, introduced Honeywell ThermoRebellion. This AI-powered solution performs the workers’ temperature screening remotely. The system was installed on the entryways of Honeywell’s US manufacturing facilities. However, it was not the first solution designed by the company. Honeywell ThermoRebellion enhanced the existing IoT technology system Honeywell Healthy Buildings, which uses connected devices and sensors to gather data on temperature, humidity, light, etc. In combination, these systems provide the full health picture of a smart building.
Implementing AI in manufacturing: a step-by-step guide
Your AI strategy must opt for top-tier technologies that align with your organization’s unique strengths and weaknesses.
The most common expectations businesses have from AI are cost reduction, process optimization, and process automation. However, to use the capabilities of AI in manufacturing industry to the full extent, organizations have to prepare for its integration. Deloitte’s survey on AI adoption summed up the key steps manufacturing businesses are planning to take in the near future as a part of their AI strategy:
How can manufacturers prepare for AI adoption?
If you find yourself unsure where to start your AI adoption strategy, here is a set of questions that will help you gain clarity and direction:
- What are your primary business goals for adopting AI technology?
- How do you envision AI contributing to your company’s growth and success?
- Do you have a manual process that AI can replace?
- Do you have a process that can be validated and enforced by AI to improve the quality and reduce defects?
- Do you have access to high-quality, structured, and relevant data for AI training?
- Have you identified potential data sources and data integration challenges?
- Do you have a team with AI expertise, including data scientists and machine learning engineers?
- Are your existing IT professionals equipped to support AI implementation?
Integration with Existing Systems:
- Have you assessed how AI solutions will integrate with your current IT infrastructure and software?
- Are there any compatibility or integration challenges to be addressed?
Scalability and Future Growth:
- Have you considered how easily your AI solutions can scale as your business grows?
- Is your infrastructure capable of handling increased AI workload and demand?
Start your AI journey with Euristiq
The Euristiq team possesses the knowledge and capabilities to guide your organization on the most suitable AI solutions that align with your business strategy. We can analyze your operations and identify critical areas where AI can drive significant improvement, be it predictive maintenance or quality control. Our comprehensive understanding of the Internet of Things (IoT) complements these strengths, enabling us to integrate AI with your existing IoT infrastructure. Combined with cloud adoption and legacy modernization expertise, we know how to help manufacturing businesses harness the transformative power of AI to achieve operational excellence and sustainable growth.
How is AI used in manufacturing?
What are the benefits of using AI in F&B manufacturing?
What are some concerns about AI in the manufacturing industry?
How will AI increase jobs in manufacturing?