How to Use Artificial Intelligence to Improve Quality Control
Interested in how to use artificial intelligence to improve quality control?
This is a great question an in excellent market that still is in its infancy.
According to a study, “AI is poised to reach over US$116.4 Billion by the year 2025, Deep Learning will bring in healthy gains adding significant momentum to global growth”.
Besides the huge sums to be made, innovating in this industry represents a chance to make a positive impact on people’s quality of life. Here’re a few amazing case studies of companies who hired DevTeam.Space to build their software products:
- Air Sign – Machine Learning Program For Air Signature Recognition
- Face, Sex, Age, Recognition System – Machine Learning Program
- Hit Factor – Machine Learning Image Recognition App
Artificial intelligence is everywhere. Futurologists, scientists, entrepreneurs, editors of popular scientific journals, ordinary citizens, scientists and the public all talk about it.
In this blog post, aimed at developers, CTOs and business leaders, we look at information systems, how can be used in a planning and production context and how developments in artificial intelligence have led onto a 4th Industrial revolution. However, it’s worth mentioning that despite having a plethora of amazing benefits, this advanced tech also has some downsides, so it’s a good idea to read more about its pros and cons.
We also explore how to improve quality control with artificial intelligence in a variety of industries.
For reference, here are the main areas we will cover:
- Information Systems, Planning and Resource Management
- The 4th Industrial Revolution
- AI and Wine Production
- Defect Recognition using IBM Watson
- Neural Networks
- What does the future look like?
Information Systems, planning and resource management
Information systems are a key factor for a business to be successful and help to ensure quality assurance levels are being met in businesses, manufacturing plants and production lines. It can‘t be denied, the rapid development of the IT system over the past few decades has helped increase productivity and to drive business efficiencies.
Integrated information systems are at the heart of these information systems and allow businesses to monitor all data relevant to the business which can include, but is not limited to:
• preparatory data
• manufacturing data
• service process data
• asset management data
• material and resource management data
• product management data
All of which can helps businesses lower production and quality control costs and with profit margins being a core KPI, one can understand why introducing artificial intelligence components to existing information systems is a direction the business is looking to move.
For example, consider the resource management of materials, tools, and documentation. This practice aims to ensure the right production processes yield the best cost for said resources. A solution, powered by artificial intelligence, is a natural fit for such a use case.
It can be trained with historical entities, attributes, processes, and decisions to produce decision trees with the most optimal planning and production processes which can then be presented for humans to execute.
Another example may be the generation of a Bill of Materials (BOM) which details each component for a given product that must be manufactured. AI can generate the most optimal steps required to produce the product whilst adhering to strict manufacturing standards.
Production Control. Planning and strategy
There are three basic modules of planning and production control which are:
• MSPLAN (Master Schedule Plan)
• MRPLAN (Material Requirement Planning)
• CRPLAN (Capacity Requirements Planning and Scheduling)
A functional production system needs to blend all three modules holistically to ensure successful manufacturing and quality control. Many difficulties in achieving production plans are due to insufficient representation of activities that are not part of the production process but are closely related to it. Understanding this, and with existing knowledge, new principles such as DIKW (Data, Information, Knowledge, Wisdom) can be adopted.
DIKW is the logical structure of data, information, knowledge, and wisdom completing information hierarchy where each level adds certain properties above and below the previous one.
Data is the basic level, information gives context, knowledge adds to its use, and wisdom adds when and why to use it. The DIKW model assumes:
• the data is in the form of unforeseen observation and dimension,
• information is formed by the analysis of connections and relationship between data,
• knowledge is formed using information for action
• wisdom is shaped using knowledge.
Structuring data in this logical fashion paves the way for businesses to implement “expert systems“ that go beyond traditional data manipulation and LOB (line of business applications).
Experts, typically in STEM disciplines, program expert systems with the specifics of their problem domain (data, processes, entities, and attributes) and thanks to their computation speeds and precision, the expert system is then able to surface actionable data that a human might have missed such as manufacturing anomalies.
An AI system is a departure from statically programmed expert systems and can be trained with historical data to make decisions by itself with no further programming and we‘ll look at some examples of how artificial intelligence is being applied to help optimize the production process and help increase productivity for businesses.
The Fourth Industrial Revolution
After several decades of development, today’s robots and AI systems are the leaders of the Industry 4.0 concept.
They run home-based businesses, manage home appliances, connect with smartphones, provide information at stations, airports, and tourist zones, used as a part of traffic monitoring systems capable of marking almost all kinds of vehicles, they deliver packages. They are almost completely autonomous.
The use of robots being implemented in manufacturing positions is gradually increasing, and with more companies deciding to take such a move, this means fewer workplaces for humans. Understandably, this is causing some concern among the human population and it is estimated that tens of millions of jobs will be lost due to automation implementation in manufacturing processes.
Considering that robots and AI systems significantly save financial resources and work faster and more efficiently than humans, we can say that the “right to work” might could be jeopardized. Moreover, their production is growing, due to the increasing number of uses, but also because of the demand. China is the world’s largest robot maker and their industry is increasingly being based on robots with a goal of 260.00 units produced annually after 2025.
Elon Musk and other successful entrepreneurs have recently spoken of the need for universal income to help mitigate such issues in the future.
Artificial intelligence is the ability to deal with new situations, data, and circumstances that have not originally been anticipated. When we add experience learning, which is enhanced by the constant repetition of tasks and performance measurement (machine learning), then we get an almost autonomous system.
This is something that is increasingly being deployed in robotics and in the automotive industry, in the manufacturing process itself, but also in HR departments, wine production, pharmaceutical industry and various services sectors, with the purpose, among others, of considerable cost savings through proper performance targeted consumption of energy and other resources and enhanced quality control.
Wine will, as it seems, in the future be made by AI systems. Australian company Ailytic has developed an AI-powered system that will helps large wine companies to optimize their production process. The technique used in production optimization has been called “prescriptive analytics” and its result is complete automation of temperature control, inventories and many other aspects of wine production.
Based on the set parameters, the program, and its algorithms create the best possible layout of the production steps which all helps drive business savings.
Defect recognition with IBM Watson
Hire expert developers for your next project
1,200 top developers
us over the last 3 years
IBM is no stranger to AI and the IoT (Internet of Things). The tech giant built a solution around their Watson platform that provides a service that can inspect physical products.
The idea was to provide manufacturers “cognitive assistants” in order to reduce errors and maximize product quality. At an early stage of production, during testing, Watson shortened the validation time by 80 percent, and errors in manufacturing fell by 7 to 10 percent.
There are several components in the solution, you define the range of “training” images and store them in Watson. Watson then concludes whether the new products are “good” or “defective” using its training data. It can also be set up on different production lines, can be trained onsite and auto-suggest changes which make the system smarter to further develop the inspection process
The system uses an ultrasonic resolution camera that captures the products while they are moving through production and assembly. Together with inspectors, Watson then recognizes defects in products, including scratches and damage.
This type of technology can be applied to a variety of industries, including the electronics and automotive industries, said Bret Greenstein, IBM Global VP for IoT.
Reduced Downtime and Increased Productivity
According to IBM data, more than half of product quality checks include some form of visual confirmation. The visual inspection ensures that all parts are in the correct position, have the right shape, color, and texture, and have no scratches, holes or side particles. Automation of visual inspection is difficult to accomplish due to the quantity and variety of products. Especially if one considers that the mistakes can be of any sort and size.
The system continually learns based on human evaluation and validation to spot mistakes in the images. The tool is designed to help manufacturers achieve the level of specialization they could not achieve so far, improving quality control, reducing downtime and increasing productivity.
You can find out more about IBMs IoT projects here.
Siemens sees predictive maintenance as being another key area where artificial intelligence can save businesses money.
Predictive maintenance algorithms can monitor components at given timeframes, based on configurable parameters and components attributes. Algorithms can then notify humans when the algorithm is 80% certain that a critical component will fail within the next 12 hours, all of which can help improve the quality of a manufacturing process.
You can find out more about Siemens Digital Factory Division here.
Artificial Neural Networks
An artificial neural network, in its simplest explanation, is a powerful machine learning model that can be used to solve complex business problems. The model is an attempt by computer scientists to mirror the structural models of the human brain and as such, are capable of processing nonlinear relationships between multiple sets of inputs and outputs.
Just like the human brain, an artificial neural network is modeled around the concept of neurons that interact with each other through an interface which consist of axon terminals which are connected to dendrites across a gap (synapse) which you can see in the image below:
Or in plain English, neural networks can make multiple decisions in parallel (whilst taking all data points into account) just like a human does.
In the case of quality control, a key advantage of neural networks is their flexibility or the possibility of finding adequate solutions based on incomplete data. Artificial neural networks can also learn from examples and training data sets so they can quickly crack problems that are often difficult to solve by more traditional software programs.
Tasks that were once performed by humans can now be passed onto a neural network for processing, thereby freeing up the workforce to focus on other activities that machines still struggle with. A single trained network can also successfully control a production system in conditions where problems arise, again, relieving the workforce of mundane, manual maintenance tasks.
These types of artificial intelligence solutions can be found in the petroleum industry, distribution of electricity, and it is even being tested for airplane control. Some applications of neural networks include, within the context of manufacturing include, but are not limited to the following:
• process control,
• product design and analysis
• quality inspection systems
• machine maintenance analysis
What will Quality Control look like in the future?
Foxconn, a technology manufacturer is often contacted by various firms to assemble their products, some of their clients include businesses such as Apple and Microsoft.
Back in 2015, in a bid to improve working conditions for its human workforce, the firm made a commitment to deploy 10,000 industrial robots to help deal with the ever-growing manufacturing demands for populate products such as the iPhone and X-Box.
According to Foxconn, the process has been slow and whilst automating simple repetitive tasks is one thing, introducing systems that leverage AI is a different thing altogether. The firm found that AI system still does not have the cognitive ability of a human worker who is much better in the final phases of quality control and aesthetics checking products.
It stands to reason that in the future, whilst some jobs may be lost to automation, others will be created as human capabilities are augmented with the arrival of AI-powered solutions.
In this blog post, we’ve explored information systems, and how advancements in production and planning strategy, coupled with advancements in software development and expert systems have led onto the creation of cognitive computing solutions that can improve the manufacturing and quality control process.
In the next blog post, we‘ll dive into 10 examples of artificial intelligence, specifically Machine Vision, that can be used in the manufacturing process.
Frequently Asked Questions
All systems can be improved by integrating new technologies and approaches. In the case of quality control, the implementation of AI and blockchain supply chain solutions will dramatically boost quality control.
We can improve AI by constantly developing and refining the code it is built on. As developers continue to improve the underlying code, AI systems will themselves be improving as they learn from their past results and errors.
The vast majority of the process that we all rely on today can be improved using AI. One day, AI will be capable of very advanced mathematics which will allow it to solve problems that have bewildered humans for years.