Introduction to Industrial Artificial Intelligence
The systematic implementation of Industrial Artificial Intelligence enables secure data sharing among manufacturers, remote monitoring and operations, managing geographically distributed assets and more.
Advanced technologies and methodologies based on Artificial Intelligence (AI) and machine learning have long been considered too academic and unreliable. It’s also been viewed as a technology that would require too many infrastructure changes, too drastic a cultural shift and too great an investment to implement. More importantly, there is often a lack of compelling evidence to convince the industry that AI technology will work repeatedly and consistently with a return on investment.
In my recent book, Industrial AI: Applications with Sustainable Performance, I directly address these challenges by introducing a concept of Industrial AI (IAI), which is a systematic discipline with standardized data processing procedures that delivers consistent and reliable solutions for industrial applications. This concept both encompasses and builds upon the pioneering work conducted at the Center for Intelligent Maintenance Systems (IMS) over the past 20 years. It exploits the convergence of real-time sensing, machine learning, advanced cyber infrastructures and decision-making tools to improve the agility, resiliency and reliability of manufacturing systems and industrial assets. The systematic implementation of IAI enables secure data sharing among manufacturers, remote monitoring and operations, managing geographically distributed assets and more. Thus, it makes its real impact on the next generation of industrial systems.
Systemization and standardization are the primary reasons why IAI is more reliable and consistent than traditional AI. Systemization of technology is created by establishing a unified framework with standard protocols to regulate model development by collectively considering the relationship among data, model and task. This framework provides clear answers to fundamental questions such as: Given a specific task, what is the optimal plan to generate and collect data from an industrial system? Given a training data set, what is the best algorithm to build a model? Given a trained model, what additional data will bring the most significant improvement to the model?
In addition to standard protocols for model development, IAI also encompasses the standardization of data acquisition, model construction, fault-tolerant mechanisms and prediction-based operating procedures and more. Without systemization and standardization in IAI, algorithms that are developed in the general domain of AI may perform poorly when directly applied to the industrial systems.
Manufacturing resilience is one of the key benefits of IAI that empowers manufacturing systems with sustainable performance.
By translating real-time sensory data into equipment health through the use of AI algorithms, predictive analytics can be leveraged to provide insights into future equipment performance and estimate the remaining time before failure. This knowledge will enable manufacturers to decide the optimal time window to plan maintenance activities so that unexpected downtime is avoided and productivity is unaffected.
The IMS Center at the University of Cincinnati piloted just such an approach with one of its member companies, a band saw manufacturer. Data from vibration sensors, acoustic sensors and controllers are merged to predict the remaining useful life of the saw blade in the developed technology. This information is accessible to the teams responsible for spare parts inventory and maintenance planning, enabling them to take corrective actions before the degraded saw blade can damage the workpiece. Currently, the band saw OEM has been taking advantage of this technology to enlarge its market share.
Data-driven adaptation of manufacturing processes has demonstrated great ability to enhance manufacturing precision, alleviate metrology burdens and shorten ramp-up time. For example, in the semiconductor manufacturing processes, the controllers in most processes (such as etching, chemical mechanical polishing, physical vapor deposition and more) are often tailored by experience to fit quality targets at great cost. However, with the knowledge learned from the AI models or the knowledge transferred from other similar processes, the learning period and expense for process control can be significantly reduced. The AI model can also be deployed to measure the part quality virtually based on the process parameters from the equipment. As a result, the controller can be adjusted once anomalies are identified, reducing the need for costly metrology devices. More importantly, the feedback mechanism in the advanced process control loop can effectively reduce process uncertainties and improve manufacturing precision.
In summary, Industrial AI is a systematic discipline that focuses on developing, validating and deploying various machine learning systemically and rapidly for industrial applications with sustainable performance. Successful implementation of technology can significantly improve agility, resiliency, and reliability of manufacturing systems and pave the way for industry 4.0.
About the Author
Jay Lee is the founding director of the Intelligent Maintenance Systems (IMS) Center and Industrial AI Center at the University of Cincinnati. Visit imscenter.net.
Selecting the right bar feeder can be one of the most important decisions a shop can make. Bar feed systems help improve productivity, throughput and quality, but in order to achieve the most benefit from them it is essential that a bar feed system be matched to the particular needs of the turning operation.
Some primary factors are often overlooked when considering how to justify the implementation of a bar feeder for turning operations.
Read about how you can automate the loading/unloading of your CNC lathes to reduce the labor cost on longer running jobs.