TMCnet News
The Internet of Manufacturing Things [Industrial Engineer](Industrial Engineer Via Acquire Media NewsEdge) Awareness in sensors, standards and software could connect virtually everything. could coiniect virtually everything Internet of things - n. the interconnection of sensors, devices, unci everyday objects Internet of Manufacturing Things - n. how the industrial Internet adds value to the manufacturingprocess THERE HAS BEEN MUCH DISCUSSION about the Interner of Things (IoT) and the Industrial Internet ot Things. The first definition above may seem vague, but it suits the examination and application in the manufacturing world. It simply means that everything, not just computer and data networks, can send and receive data. IoT can improve various aspects of manufacturing efficiency' including productivity, asset health, profitability, quality, safety-', employee safety and environmental impact. The narrow slice ot the Industrial Internet of Things that mostly affects industrial engineers is the Internet of Manufacturing Things. The critical relevance is the specific focus on how the industrial Internet adds value to the manufacturing process. Motivation Until recently, the manufacturing sector has been underserved with the new technology developed in the IoT space. The sector has great potential for productivity improvement and value engineering. Much of the extraordinary volume of data generated by manufacturing equipment falls to the ground, uncollected for downstream analysis. The loss of the data fritters away the opportunity' to mine it for information and find ways to improve productivity. The command and control loop in a modem multiple-axis machine tool offers insight to how information is regularlylost. The data buses inside the machine tool transport large volumes of data at high speeds between the controller, the various actuators and motors, and the humanmachine interface when the machine is in operation. Very little of these data are captured and managed to support decisionmaking about the machining process. The operation of the current manufacturing economy poses interesting challenges in accomplishing an overall improvement in manufacturing efficiencies. Global manufacturing today is fragmented and heterogeneous. Manufacturing systems are located across large geographic areas, and a single product might require patts and assemblies made in plants from multiple states and countries. These manufacturing supply chains also are highly fragmented. No single company controls the entire supply chain; all manufacturers are dependent on a network of suppl iers who al I pa rtici pate i n th e procèss of adding value to the product. This global reach and supply chain diversity' ensure that manufacturing operations looking for improvement need dynamic methods, ideas beyond a business-as-usual solution. The exponential increase of sensors and devices that generate data creates new approaches to interrogate and understand what is happening with manufacturing equipment. Simultaneously, the growing complexity of global manufacturing systems requires holistic solutions, which is where the Internet of Manufacturing Things can enter the conversation. Process traceability Tying together the complexity-- of global manufacturing systems with a holistic approach is challenging and requires that systems are da-eloped for process traceability. Process traceability must track and trace aery' process that happened to a part as it went through the manufacturing process. Industrial engineers are keenly aware of the distinction between part traceability and process traceability. Part traceability primarily deals with what part was manufactured, along with when and where. Process traceability builds and expands on this information to understand fully' how the part was manufactured. Process traceability' is a key' requirement in aerospace, medical and other high-precision manufacturing sectors, since manufacturing defects have sa'ere (and life-threatening) impacts on the quality' of parts. With capable process traceability systems, illustrated in figure i, industrial engineers can go back in ti me and find out exactly how and when the manufacturing defect was introduced. At best, most manufacturing companies have part traceability systems reporting only when which patt was manufactured. When automated, this type of data is collected by the manufacturing execution system or enterprise resource planning (HRI') systems. With effective aggregation and rollups, lire data reveal when a particular batch of parts was manufactured, along with the lot number. This is insufficient data; it is nor enough to understand fully what happened when a part was made. To build a full-fledged process traceability' system, the following data are needed: * Identity data are the most basic kind required for part traceability. This kind of data reveals what is being made, how much was made and when it was made. Examples include the heat ID, batch ID and operator ID. * Operational data establish what the machine was doing when it had a pait associated with its operation. The data allow industrial engineers to understand the utilization of a device when it was operating on a part, how long the device was in "auto" mode versus "manual" inode, and the different downtimes that device experienced when working on the pait. Knowing the downtimes indicates potential issues with part quality. When a device had repeated unplanned maintenance downtimes when it was producing a part, there is great likelihood that the pari has some quality problems and might require additional metrology. Examples include device uptime, downtime, modes and stares. * Diagnostic data further elaborate the correlations started using operational data. Alarm and condition data can reveal specific issues in the device that manufactured a part. Examples include alarms, warnings, messages and notifications. * Process data generate significantly more detail and reveal how specific features were generated on the part. High-speed machining of aerospace alloys, for example, requires preservation of a specific chip velocity' when features are being created. But interpolation errors and machine tool limitations often result in significant variations between the actual feed rate and the planned feed rate. With process data, manufacturers know exactly when these deviations happened and understand the impact. Examples include positions, velocities, acceleration and flow rates. * Environmental data establish the impact of the part as it is being manufactured. With detailed knowledge of the resource Hows associated with the part, manufacturers accurately estimate environmental impact. A high level of detail reveals which stages of the production process have the greatest impact; that knowledge targets energy efficiency improvements. Examples include resource usage, energy consumption, effluents and emissions. "With a process traceability system, analytics are built to fully understand the manufacturing process execution and its impacts and apply this information in improving efficiencies," according to Athulan Vijayaraghavan, chief technical officer with System Insights. "This kind of a system can also help in integrating with upstream applications that look at the design of the pait and downstream applications that look at the use-phase of the part." Solving the challenge The Internet of Manufacturing Things needs the following three key pieces to help solve tli is grand challenge of process traceability: standards, sensors and software. Standards are vital in building the Internet oí ManufacturingThings. Manufacturing data is highly complex, and the specialized technical knowledge needed to work in the domain acts as a deterrent to innovation and new ideas from the outside. Manufacturing also operates as islands of excellence; innovation occurs in specialized areas, and improvements often are not shared across tine wider enterprise. Standards repair these nonintegrated behaviors by providing a transparent way to exchange information and increase visibility across these islands. This is diagrammed in Figure 2. One standard that is proving effective in bridging these islands is MTConnect, a data exchange standard that allows for disparate entities in a manufacturing system to share data seamlessly in a common format. MTConnect is an open, royalty-tree standard that describes data from manufacturing equipment using a common, unambiguous vocabulary. The MTConnect standard models the manufacturing domain, allowing devices, equipment and systems to output data in an understandable formal that can be read by any other device using the format. The standard is based on XML (extensible markup language), which offers a widely recognized and accepted flexible representation for exchanging semi-structured machine-readable data. This approach allows connectivity from the lowest end of the process chain, nearest the work piece or shop floor, to the highest design or process-planning tool. Additionally, the interoperability provided by MTConnect enables a host of third-party solution providers to develop software and hardware to make the entire manufacturing enterprise more productive. MTConnect can address a wide variety of data, including physical and actual device design data, measurement or cali bration data and near-real-time data from the device. Sensors are needed to measure what happens in the manufacturing process to enable decision-making and automation. Sensors are seen at every level in the manufacturing system, from the individual process throughout the supply chain. In manufacturing, physics-based sensors must make measurements based on fundamental process technologies. Sensors must be minimally invasive and capture measurements without interrupting or influencing the phenomena being measured. Sensors should support data communication using open standards so that they can operate in the wider system. Sensor-based systems drive manufacturers to ask where to put the intelligence to manipulate all of the sensor data. The traditional approach has been to localize sensor intelligence and create self-contained local command and control loops. However, this approach isolates the sensor loops, which then share no data and provide no way to paint a bigger picture. A tempting possibility'when considering networked manufacturing is centralized intelligence. While centralized decisionmaking brings a more holistic approach, the challenges in data loads,limited bandwidth and latency surface quickly. A solution to this problem is creating distributed decentralized systems, which splits decision-making between local and centralized controllers. Based on available computing cycles and bandwidth, decisions can be moved to different locations. Software applications are required to manage the data generated trom the Internet of Manufacturing Things and apply these data in decision-making. The size of your manufacturing system will decide what kind of data loads your software and hardware must be capable of handling. Conservative estimates based on data generated from manufacturing devices and sensors in different types of manufacturing facilities show that a small facility could generate 2 to to terabytes a year, a medium facility 5 to 25 terabytes a year, and a large one 16 to 80 terabytes a year. An entire enterprise would need to handle 80 to 5,000 terabytes a year, while theentire U.S. machining sector could generate from 200 petabytes a year to 1 xylobabyte a year. To put in perspective just how much data we are talking about. Wikipedia defines a terabyte as 1,024 gigabytes, a petabyte as 1,024 terabytes, and a xylobabyte is to to the 36th power (1036). "Dealing with these data loads is not trivial," Vijayaraghavan noted. "It requires software to capture, store and transfer the data without losing any detail." Software also is needed to manage the different types of data generated across manufacturing systems.This data generally falls into three types: structured data, which includes data pertaining to the process parameters from sensors and telemetric systems; unstructured data, which includes data pertaining to alarms, faults, quality control, performance and tests; and tribal knowledge, which includes data captured from skilled and semi-skilled operators of the manufacturing equipment, along with behavioral inputs in the form of annotations, overrides and interruptions Making the decision Decision-making software for manufacturing must encompass several core considerations, such as reason over multiple events, reason over multidimensional hierarchical data, and reason over temporal scales. * Reason over multiple events: Manufacturing data can be modeled as a stream of events. These events are analyzed to understand and track the behavior of the manufacturing equipment. Event processing requires capabilities to fuse events from multiple streams and create abstract complex events, filter events based on static and dynamic parameters, aggregate events and compute metrics, and identify relationships between events. Relationships can be evaluated using temporal and spatial patterns, and the results of these evaluations are used in generating events. * Reason over multidimensional hierarchical data: Manufacturing data also are multidimensional and can be modeled in terms of multiple orthogonal hierarchical axes, including equipment hierarchy that identifies the device, line, cell or factory; part hierarchy that identifies the part, assembly or product family; and operator hierarchy that identifies the operator or the work group. Software must capture and model diese multidimensional hierarchies to allow' for reasoning across these multiple dimensions of time, machi ne organization and parts. * Reason over temporal scales: A special case of hierarchical reasoning is reasoning over multiple temporal scales. Manufacturing decisions occur across temporal scales ranging from microseconds at the level of the process interface, to seconds and minutes at the level of process planning, to days and months at the level of enterprise planning. Software for the Internet of Manufacturing Tilings must be able to deal with decision-making across all these scales. Putting it all together So 10 sum up, 10 have an effective and sustainable Internet of Manufacturing Things, enterprises need sensors to capture high-speed dara from heterogeneous sources and to transmit this data in standardized formats, standards that are robust and flexible to enable integration across multiple hardware and software platforms, and software for decision-making across spatial and temporal resolutions. Tremendous advancements have been made in the areas of sensors, standards and software to support these generic requirements, hut more work is needed in developing capabilities to support the specific requirements of the manufacturing domain.This demand for innovation is being met by many small and large companies that address niche elements ot the industry sector. The strategic researchers at Harbor Research suggest that vertical expertise will become an increasingly critical differentiator, yet diet' question whether large IT vendors and carriers that have dominant shares and scale economies in horizontal technologies will add new ''vertical" solution elements to their infrastructure offerings. Time will determine if smaller vendors focused on vertical niches perform in a w'orld increasingly dominated by behemoths. The coming years should yield extensive activity and product development to realize the vision of die Internet of Manufacturing Things, a space that could be valued in the tens of billions of dollars by2020. 'MIND-BOGGLING' NUMBERS According to Business Insider, the volume of devices associated with the Internet of Things (IoT) will be "mind-boggling." By 2018, 9 billion different devices that range from parking meters to home thermostats will be connected. It will drive trillions in economic value as it permeates consumer and business life. IoT products already include kitchen and home appliances, lighting and heating products and car monitoring devices that let motorists pay insurance only for the miles they actually drive, according to the media company. Industrial uses will include Internet-managed assembly lines, connected factories, warehouses and supply chains. Electricity grids will be able to adjust rates for peak energy usage. Cities like Cincinnati, Ohio, already have seen residential waste decline through its "pay as you throw" program that uses IoT technology to monitor waste. And several cities have reduced water leaks by 40 percent to 50 percent by putting sensors on pumps and other infrastructure. Thomas R. Cutler is the president and CEO ofTR Cutler Inc, and the founder of the Manufacturing Media Consortium, which includes more than 5,000journalists, editors and economists uniting about trends in manufacturing, industry, material handling and process improvement. Cutler writes more ihan y 00 failure (Micks annually about the manufacturing sector. (c) 2014 Institute of Industrial Engineers-Publisher |