Machine to machine
Machine to machine (M2M) is direct communication between devices using any communications channel, including wired and wireless.[1][2] Machine to machine communication can include industrial instrumentation, enabling a sensor or meter to communicate the information it records (such as temperature, inventory level, etc.) to application software that can use it (for example, adjusting an industrial process based on temperature or placing orders to replenish inventory).[3] Such communication was originally accomplished by having a remote network of machines relay information back to a central hub for analysis, which would then be rerouted into a system like a personal computer.[4]
More recent machine to machine communication has changed into a system of networks that transmits data to personal appliances. The expansion of IP networks around the world has made machine to machine communication quicker and easier while using less power.[5] These networks also allow new business opportunities for consumers and suppliers.[6]
Machine to machine wireless networks can serve to improve the production and efficiency of machines, to enhance the reliability and safety of complex systems, and to promote the life-cycle management for key assets and products. By applying Prognostic and Health Management (PHM) techniques in machine networks, the following goals can be achieved or improved:
The application of intelligent analysis tools and Device-to-Business (D2B) TM informatics platform form the basis of e-maintenance machine network that can lead to near-zero downtime performance of machines and systems.[34] The e-maintenance machine network provides integration between the factory floor system and e-business system, and thus enables the real time decision making in terms of near-zero downtime, reducing uncertainties and improved system performance.[35] In addition, with the help of highly interconnected machine networks and advance intelligent analysis tools, several novel maintenance types are made possible nowadays. For instance, the distant maintenance without dispatching engineers on-site, the online maintenance without shutting down the operating machines or systems, and the predictive maintenance before a machine failure become catastrophic. All these benefits of e-maintenance machine network add up improve the maintenance efficiency and transparency significantly.
As described in,[36] The framework of e-maintenance machine network consists of sensors, data acquisition system, communication network, analytic agents, decision-making support knowledge base, information synchronization interface and e-business system for decision making. Initially, the sensors, controllers and operators with data acquisition are used to collect the raw data from equipment and send it out to Data Transformation Layer automatically via internet or intranet. The Data Transform Layer then employs signal processing tools and feature extraction methods to convert the raw data into useful information. This converted information often carries rich information about the reliability and availability of machines or system and is more agreeable for intelligent analysis tools to perform subsequent process. The Synchronization Module and Intelligent Tools comprise the major processing power of the e-maintenance machine network and provide optimization, prediction, clustering, classification, bench-marking and so on. The results from this module can then be synchronized and shared with the e-business system on for decision making. In real application, the synchronization module will provide connection with other departments at the decision making level, like Enterprise Resource Planning (ERP), Customer Relation Management (CRM) and Supply Chain Management (SCM).
Another application of machine to machine network is in the health management for a fleet of similar machines using clustering approach. This method was introduced to address the challenge of developing fault detection models for applications with non-stationary operating regimes or with incomplete data. The overall methodology consists of two stages: 1) Fleet Clustering to group similar machines for sound comparison; 2) Local Cluster Fault Detection to evaluate the similarity of individual machines to the fleet features. The purpose of fleet clustering is to aggregate working units with similar configurations or working conditions into a group for sound comparison and subsequently create local fault detection models when global models cannot be established. Within the framework of peer to peer comparison methodology, the machine to machine network is crucial to ensure the instantaneous information share between different working units and thus form the basis of fleet level health management technology.
The fleet level health management using clustering approach was patented for its application in wind turbine health monitoring[37] after validated in a wind turbine fleet of three distributed wind farms.[38] Different with other industrial devices with fixed or static regimes, wind turbine's operating condition is greatly dictated by wind speed and other ambient factors. Even though the multi-modeling methodology can be applicable in this scenario, the number of wind turbines in a wind farm is almost infinite and may not present itself as a practical solution. Instead, by leveraging on data generated from other similar turbines in the network, this problem can be properly solved and local fault detection models can be effective built. The results of wind turbine fleet level health management reported in[37][39] demonstrated the effectiveness of applying a cluster-based fault detection methodology in the wind turbine networks.
Fault detection for a horde of industrial robots experiences similar difficulties as lack of fault detection models and dynamic operating condition. Industrial robots are crucial in automotive manufacturing and perform different tasks as welding, material handling, painting, etc. In this scenario, robotic maintenance becomes critical to ensure continuous production and avoid downtime. Historically, the fault detection models for all the industrial robots are trained similarly. Critical model parameters like training samples, components, and alarming limits are set the same for all the units regardless of their different functionalities. Even though these identical fault detection models can effectively identify faults sometimes, numerous false alarms discourage users from trusting the reliability of the system. However, within a machine network, industrial robots with similar tasks or working regimes can be group together; the abnormal units in a cluster can then be prioritized for maintenance via training based or instantaneous comparison. This peer to peer comparison methodology inside a machine network could improve the fault detection accuracy significantly.[38]