Recent Advances and Trends in Predictive Manufacturing in Industry 4.0 Environment
Essential as it is, the U.S. manufacturing industry is experiencing challenges that threaten its sustainability. Despite claims of a rebound from the last economic recession of 2008-2010, the U.S. market share in the global export of advance technology products still dropped by six percent in 20113 and only a small portion of the five million jobs that were lost has been recovered. Furthermore, the U.S. manufacturing industry faces strong and increasingly capable competition from emerging economies that are positioning themselves to lead markets in high value-added products. Other nations have invested to strengthen their manufacturing competitiveness with intensive research and development (R&D) efforts4 more aggressively than the U.S. A report5 by the National Science and Technology Council recognizes this gap between R&D activities and the deployment of technological innovations in domestic manufacturing. The Obama administration has proposed a $1 billion investment in a National Network for Manufacturing Innovation to address the aforementioned gap6. However, the question remains: What innovative technologies should be developed in order to sustain the competitiveness and leadership of the U.S. manufacturing industry?
The globalization of the world’s economies is a challenge to the local sector and it is pushing the manufacturing industry to the brink of its next transformation. This metamorphosis is aided by current advances in sensing, instrumentation, automation, communication and other emerging technologies. This article describes the evolution of manufacturing, and introduces the concepts and principles of predictive manufacturing and how it can mold next generation production assets and their auxiliary systems to improve manufacturing competitiveness.
EVOLUTION OF MANUFACTURING STRATEGIES
From the early adoption of mechanical systems to support production processes, to today’s highly automated assembly lines, manufacturing has always been a vibrant industry with an ecosystem that is highly reliant on innovation and ingenuity. The discovery of new technologies has enabled numerous transformations and technological developments to occur (and will continuously be expected) in order to be responsive and adaptive to dynamic market requirements and demands.
At the turn of the 20th century, mass production was made popular by Henry Ford to generate large volumes of standardized units using assembly lines designed for the systematic and sequential organization of laborers, machines and parts. Production costs are minimized by utilizing interchangeable parts to build the final product. However, the early implementations of mass production were laden with numerous sources of waste, such as overproduction, waiting (idle time), transport, processing, inventory, redundant or unnecessary motion and defects. Then came the Toyota Production System in the early 1970s, the prime directive of which is to reduce production costs by minimizing and ultimately avoiding the sources of waste. This manufacturing strategy entails an organizational-wide operation evaluation to determine the different sources of waste and identify necessary changes to processes in order to avoid them. Such philosophy eventually trickled into the Western nations with the development of lean principles and Six Sigma techniques. Then, the start of the computer age brought forth numeric controllers that provided automation and flexibility to equipment, such as industrial robots, machine tools and other engineering assets. With such capabilities, manufacturers are able to institute mass customization or agile manufacturing and supply consumers with products they want. The last two decades saw astounding growth in the advancement and adoption of information technology and even social media networks that have increasingly influenced consumers’ perception on product innovation, quality, variety and speed of delivery. This scenario led to the establishment of reconfigurable manufacturing, wherein a plant structure can be easily and systematically modified so production capacity can scale up rapidly and functionality can adapt more quickly.
Manufacturing organizations can only take full advantage of the aforementioned strategies if there is a good fit between the principles of the paradigm being adopted and the company’s corporate goals, and if implementing guidelines are faithfully observed. However, even a strict compliance does not ensure maximum potential benefits due to uncertainties that can be found in machines, the people and even processes.
Figure 1: Evolution of manufacturing paradigms
TRANSPARENCY – SHOWING THE UNCERTAINTIES IN MANUFACTURING
In manufacturing, there are many uncertainties that may not be quantifiable or even known to decision makers, making them unable to form sound judgments and conclusions about the efficient operation and usage of their assets (see Figure 2). These uncertainties persist both internal and external to the factory. Examples of internal uncertainties include degradation of the machining processes and the occurrence of failure events without any recognizable symptoms (component level); and variation of cycle time due to inconsistent operation, unplanned breakdown of systems and the presence of scraps and rework that may lead to difficulties in production planning and scheduling (system or production process level). Meanwhile, external uncertainties, typically stemming from product development all the way through the supply chain, can manifest as: 1) unreliable downstream capacity, 2) unpredictable variation of raw materials or parts in terms of delivery, quantity and quality, 3) market and customer demand fluctuation, 4) incomplete product design due to the lack of accurate estimation of product state during production and usage, and 5) random warranty claims and demands for replacements, etc.7
Figure 2: Illustration of manufacturing issues in a factory
The internal manufacturing issues can be further mapped into two domains: visible and invisible. Examples of visible issues include machine failure, product defects, poor cycle times, long time delays, drops in overall equipment effectiveness (OEE), etc. These are very obvious conditions and information retrieved from analyzing visible issues is primarily after the fact. On the other hand, invisible issues may occur as machine degradation, component wear, etc. These uncertainties can have adverse effects on manufacturing operations if no predictive analytics and control strategies are judiciously implemented.
In each of the domains, issues are treated in both deterministic and uncertain levels (see Figure 3). Pertaining to deterministic issues (Q3), tools using best practices and standard work are normally utilized to handle these issues systematically. Then as a potential countermeasure, companies work with their equipment suppliers to utilize new knowledge from their internal problem-solving exercises and develop technologies that will be integrated to their equipment as a value-added improvement (Q2). Meanwhile, on uncertainty issues, efforts have been made, such as in the prognostics and health management (PHM) research area, to formulate more advance predictive analytics and detect problems at an early stage (Q3). The unmet need, therefore, is to replicate what has been done in the invisible space and further define how issues are addressed from the problem-solving level to the problem avoidance level (Q4). Utilizing predictive tools and techniques will unravel many new value-creation opportunities that will exploit the new information (unknown knowledge).
Figure 3: Problem solving and avoidance in both visible and invisible spaces
What are needed then are tools and technologies that can provide transparency, which is the ability of an organization to unravel and quantify such uncertainties to determine an objective estimation of its manufacturing capability and readiness.8 The manufacturing strategies described earlier have haphazardly assumed continuous equipment availability and sustained optimal performance during its every usage, yet such assumptions do not hold true in a real factory. In order to achieve transparency in the plant, the manufacturing industry has to take the plunge and transform itself into predictive manufacturing. Such evolution requires the utilization of advance prediction tools and approaches so data that is continuously generated by factories can be systematically processed into information. This information can help explain the uncertainties, thereby allowing asset managers and process supervisors to make more “informed” decisions.
The aggressive adoption of the “Internet of Things” ideology, even in the manufacturing industry, has helped in laying the foundation of predictive manufacturing by setting the foundational structures of smart sensor networks and smart machines. The use of advance predictive tools has become more prevalent across different market segments. An area that has leveraged on such predictive analytics is prognostics and health management. It deals with the assessment of the condition of a manufacturing asset, the discovery of incipient failures and an inference of the next failure event so proactive maintenance activities can be performed to avoid catastrophic and costly machine breakdowns.
THE PREDICTIVE MANUFACTURING SYSTEM
A predictive manufacturing system provides machines and systems with “self-aware” capabilities, thereby giving greater transparency to users and ultimately avoiding potential issues concerning productivity, efficiency and safety.
The core technology of a predictive manufacturing system is the smart computational agent that contains smart software to conduct predictive modeling functionalities. The predictive analysis of equipment performance and estimation of the time to failure will reduce the impacts of these uncertainties and give users the opportunity to proactively implement mitigating or even countermeasure solutions to prevent productivity/efficiency loss in manufacturing operations.
Predictive manufacturing systems allow users transparency in operations with information, such as actual health condition, a trajectory of the equipment’s performance or degradation behavior, and insights as to when and how the equipment, or any of its components, is likely to fail.
Some benefits of a well designed and developed predictive manufacturing system are:
Cost reduction – By knowing the actual condition of the manufacturing assets, maintenance activities can be provided at a more appropriate condition (not too late that a failure has occurred and not too early that a perfectly good part is being unnecessarily replaced). This is also known as just-in-time maintenance.
Operation efficiency – With knowledge when equipment is likely to fail, production and maintenance supervisors can prudently schedule their activities, thereby maximizing equipment availability and uptime.
Product quality improvement – Degradation patterns and near real-time machine condition estimates can be integrated with process controls so product quality is maintained while accounting for equipment or system drifts over time.
PROGNOSTICS AND HEALTH MANAGEMENT
Prognostics and health management (PHM) is a critical research domain that leverages advanced predictive tools. PHM is becoming more popular primarily due to the need to have a mechanism to obtain objective assessment on the true condition of manufacturing assets, as well as auxiliary systems. In the past, reactive maintenance has been adopted widely due to its practical approach, wherein a machine is repaired only when it is needed (i.e., when it fails). However, as production throughput rates have soared to meet growing consumer demands, unplanned downtime has become prohibitively expensive and must be avoided. Then came preventive maintenance strategies that require maintenance activities (such as conditioning and replacement) to be accorded either on a time- or usage-based interval. Although the preventive maintenance approach can provide the highest level of availability (assuming a reasonable time interval), it has two major drawbacks: (1) implementing a preventive maintenance program is expensive, especially if the time interval is very short and (2) since components are being replaced before failure or even before symptoms start to show, there is no insight gained about the degradation behavior of the asset. Condition-based maintenance, meanwhile, utilizes machine signals (either from the controller or from sensor installations) to detect the occurrence of a fault or anomaly. In some implementations, even the location of the fault can be isolated and the type of fault event can be recognized. PHM is a natural extension of condition-based maintenance by using prediction algorithms so future performance of the equipment can be inferred. By monitoring health metrics (confidence value or CV, failure mode, remaining useful life, etc.), the user can observe the temporal behavior of the machine and be warned of the incipient signs of failure before the actual fault event. With such information, manufacturing transparency is then achieved because plant managers and supervisors are now capable of identifying machines that can optimally finish a production job order (mission readiness) while prioritizing equipment for repair without interfering with production schedules.
A FRAMEWORK FOR A PREDICTIVE MANUFACTURING SYSTEM AND PHM
In order to reap the benefits of a predictive manufacturing system, essential components should be present and effectively utilized as illustrated in Figure 4. It starts out with data acquisition of appropriate signals using the applicable sensor assemblies to extract data, such as vibration, temperature, pressure, electrical signals, etc. It also may be useful to apply data mining and correlation to historical data to augment the incoming data. Moreover, there are industry-grade communication protocols, such as MTConnect and OPC (OLE-DB for Process Control) that can aid users to store status signals from the machine’s controller. Such data can provide context information when the sensor data is being simultaneously recorded. When all the data from the different sources (historical, sensors and controllers), multiple components and units are aggregated, then the big data predicament needs to be addressed. Such phenomenon poses the challenge of how to effectively manage and extract useful information only. An example of an effective transforming tool to manage big data is the Watchdog Agent™, developed by the National Science Foundation’s Industry/University Collaborative Research Center (I/UCRC) for Intelligent Maintenance Systems (IMS, www.imscenter.net) in 2001. It is a suite of predictive tools and algorithms that can be categorized into four sections: signal processing and feature extraction, health assessment, performance prediction and fault diagnosis. With prudent selection and use of visualization tools, health metrics, such as current condition, remaining useful life, failure mode, etc., can be effectively conveyed in terms of radar chart, degradation curve, risk chart and health map. With integration to the manufacturer’s enterprise resource planning (ERP) system, the health metrics can trickle into the company’s other corporate wares, such as supply chain management (SCM), manufacturing execution system (MES) and customer relationship management (CRM), allowing for a more holistic enterprise control and optimization. With manufacturing transparency, management has the right information to compute facility-wide OEE. Equipment can then be managed cost-effectively with just-in-time maintenance. Finally, historical health information can be provided to equipment designers for closed-loop lifecycle redesign to improve next generation production systems.
Figure 4: Predicting a manufacturing system framework using predictive analytics
The aggressive adoption of the “Internet of Things” ideology by the manufacturing industry has ushered in the unique opportunity to unravel and quantify uncertainties in assets and processes, ultimately improving manufacturing competitiveness. A predictive manufacturing system is presented here as the next phase in the industry’s evolution that can provide transparency in the factory. Through the use of advance predictive tools and techniques, such as those found in the PHM practice, users can objectively and confidently deal with invisible uncertainties.
Bureau of Labor Statistics, U.S. Department of Labor. Industries at a Glance – Manufacturing: NAICS 31-33. http://uptime4.me/bls-gov
U.S. Census Bureau, U.S. Department of Commerce. Trade in Goods with Advance Technology Products, 2010 and 2011. http://uptime4.me/census-gov.
The President’s Council of Advisors on Science and Technology. Capturing Domestic Competitive Advantage in Advanced Manufacturing. AMP Steering Committee Report, July 2012. http://uptime4.me/whitehouse-gov.
National Science and Technology Council Committee on Technology, Interagency Working Group on Advanced Manufacturing. A National Strategic Plan for Advanced Manufacturing. February 2012, http://uptime4.me/whitehouse-gov-pdf.
The White House, Office of the Press Secretary. President Obama to Announce New Efforts to Support Manufacturing Innovation, Encourage Insourcing. March 9, 2012. http://uptime4.me/whitehouse-pres.
Lee, Jay; Lapira, Edzel; Yang, Shanhu; and Kao, Hung An. Predictive Manufacturing System Trends of Next-Generation Production Systems. In the Proceedings of the 11th IFAC Workshop on Intelligent Manufacturing Systems 11 1:150-156, 2013.
Lee, Jay and Lapira, Edzel. Predictive Factories: The Next Transformation. The Manufacturing Leadership Council, Frost & Sullivan: Manufacturing Leadership Journal, February, 2013.
Dr. Jay Lee is Ohio Eminent Scholar and L.W. Scott Alter Chair Professor at the University of Cincinnati and is founding director of National Science Foundation (NSF) Industry/University Cooperative Research Center (I/UCRC) on Intelligent Maintenance Systems. His current research focuses on dominant innovation tools for product and service design, as well as intelligent prognostics tools and smart predictive analytics for equipment reliability assessment and smart product lifecycle management.
Edzel Lapira is the associate director of the NSF Industry/ University Research Cooperative Center (I/ UCRC) for Intelligent Maintenance Systems (IMS), and a research associate at the University of Cincinnati. He has worked with several multinational companies on IMS projects, such as Caterpillar, Sinovel, Toyota, Goodyear, Nissan and National Instruments. www.imscenter.net