Edge computing keeps industries more safe. How?
The cloud is not a hard requirement for Industrial IoT. You don´t necessarily need a cloud IoT platform, at least to great extent where you can secure your production & enterprise data that is being exposed to external world. Industry 4.0 is all about connecting machines, so your manufacturing processes can react more quickly and intelligently to changing factory floor conditions. Connecting assets will help you achieve greater levels of agility and automation. But it will also increase your risks. A more connected organization is one that offers many more attack surfaces and a much higher degree of vulnerability to cyber-attacks. Your Industry 4.0 strategy can minimize the risks with the adoption of edge computing in your solutions. If as much data as possible gets processed on the edge, rather than being sent to the cloud, there is a much lower risk of it being intercepted or tampered with. Robust edge computing systems will let you keep the bulk of your IT and operational technology systems on your secure network.
Nowadays industrial machine maintenance is mainly reactive and preventive, where the predictive strategy is applied for critical situations only. Traditionally these maintenance strategies are not taking into consideration the huge amount of data being generated from machines and the available emergent Information and Communications Technology (ICT), for example: Industrial IoT, Big data analysis, advanced data analytics, cloud computing and augmented reality. However, the maintenance paradigm is changing and industrial maintenance is now understood as a strategical factor and a profit contributor to ensure good productivity in the Shopfloor. This shift in the maintenance paradigm has led to the research and development of new ways to execute machine maintenance by considering the operational state of assets and this has enabled the development of new maintenance approaches such as the Prognostic and Health Management (PHM) & the Condition Based Maintenance (CBM). In other ways, these approaches apply data analysis techniques to the information produced in the shop floor processes to detect anomalies in the machines behavior.
Thus, having in mind the improvement of the performance of the production process, this work aims to develop an intelligent and predictive approach for the industrial maintenance, aligned with the Industry 4.0 principles, that considers advanced analysis of the data collected from the shop floor to monitor and earlier detect the occurrence of disturbances and consequently the need to implement maintenance actions. This approach extends PHM and CBM maintenance approaches by considering machine learning and augmented reality technologies to support maintenance technicians during the maintenance interventions by providing a guided intelligent decision support articulated by the use of human-machine interaction technologies.
Usually, the maintenance management is categorized into different policies:
- Corrective or run-to-failure maintenance (Breakdown maintenance)
- Preventive maintenance
- Predictive maintenance
Corrective maintenance is an unscheduled repair, where equipments are allowed to operate until they fail, moment in which a maintenance intervention is performed. Preventive maintenance, probably the most popular maintenance policy, is a regularly performed set of actions on an equipment to lessen the likelihood of it failing. This type of maintenance is performed while the equipment is still working and is planned so that any required resources are available. Finally, predictive maintenance is a philosophy or attitude that uses the actual operating condition of the plant equipments and systems to optimize the plant operations and/or processes. Predictive maintenance concerns the application of sensor technology and analytical tools to predict when equipments’ failures might occur and to prevent the occurrence of the failures by performing maintenance. The failures’ prediction can be done by applying, e.g., vibration monitoring or thermography, and must be effective at predicting failures and also provide sufficient warning time for the upcoming maintenance. When this policy is working effectively, maintenance is only performed on equipments when it is required.
The PHM concept is often used with other approaches like predictive maintenance and CBM. PHM is an engineering process where algorithms are used to detect anomalies, diagnose faults and predict Remaining Useful Lifetime (RUL). Although the main goal of PHM is to provide the health state and estimate the RUL of the components or equipments, also financial benefits such as operational and maintenance cost reductions and extended lifetime are achieved. A PHM analysis involves a variety of steps including the collection of data and data characterization, the extraction of features from collected data, and finally the diagnosis and prognosis.
Typical functional blocks of such predictive maintenance system are…
- Data Acquisition: provides the access to digitized sensor or transducer data and records this data.
- Data Manipulation: may perform single and/or multichannel signal transformations and may apply specialized feature extraction algorithms to the gathered data.
- State Detection: performs condition monitoring by comparing features against expected values or operational limits and returning conditions indicators and/or alarms.
- Health Assessment: determines if the system’s health is suffering degradation by considering trends in the health history, operational status and maintenance history.
- Prognostics Assessment: projects the current health state of the asset into the future by considering an estimation of future usage profiles.
- Advisory Generation: provides recommendations related to maintenance actions and modification of the asset configuration, by considering operational history, current and future mission profiles and resource constraints.
Solution architecture approach:
A proper solution architecture for condition-based maintenance should present specific modules such as those described in the previous section. The proposed system architecture integrates all the aforementioned referred modules to create a functional system that allows the implementation of intelligent and predictive maintenance, taking advantage of a broad spectrum of technologies, such as IoT, machine learning, expert systems, among others. The developed architecture is depicted in Figure above.
The system functionality is initiated with the Data Collection module, where the data from several sources is collected and stored in a database. This database will feed the Offline Data Analysis module, where advanced data analytics, machine learning and cloud technologies are used to perform the knowledge generation. The outputs of this module are the generation or adjustment of rules, procedures and facts, which will be used by the Dynamic Monitoring functional block.
The Dynamic Monitoring module is divided into two components, the Visualization and the Early Detection of Failures. The Visualization component allows to compare Key Performance Indicators (KPIs) against the expected operational limits. In order to determine these operational limits the facts resultant from the Off-line Data Analysis are considered and the raw data is displayed in a graphic format to facilitate its interpretation. On the other hand, the Early Detection of Failures component processes the facts and rules through the use of an inference engine, and triggers a maintenance warning when an anomaly is detected in an earlier stage. Depending on the detected anomaly, the maintenance warnings can lead to different maintenance actions, namely corrective, preventive or predictive. Once the need for a maintenance intervention is detected, this information is sent to the scheduling tool that will schedule the intervention according to the current production state and the maintenance resources availabilities. In spite of the importance of the scheduling system, this is out of scope of this work and consequently will not be detailed in this paper.
The execution of scheduled maintenance interventions is guided and supported by a decision support system that selects the appropriate maintenance procedure and translates it into a language understandable by the human. The Intelligent Decision Support module is also able to adapt or create new maintenance procedures in cases that there are no known maintenance procedures for the detected anomaly. The maintenance procedure is provided to the maintenance technician, while performing the required maintenance actions, by using advanced Human-Machine Interfaces (HMI), e.g., head mounted devices.