Global enterprises are advised to use real-time monitoring, data collection, combined with back-end analysis applications in order to access early alerts, notifications and any updates to avoid unexpected equipment anomaly or breakdown, according to Bruce Lu, senior consultant, Greater China Pre-sales Support Department, SAS Institute Taiwan.
The use of numerical data - such as temperature, pressure, flow, voltage and electric current incurred during the production process - as the basis for maintenance as well as prevention of sudden malfunction of costly equipment is not enough to help enterprises reduce the risk of having unexpected failure of their equipment, Lu said.
Enterprises must return to the fundamentals to pragmatically look at the data generated by related equipment, to see what data has been ignored previously and how to use the date to prevent in advance their machines from malfunctioning, Lu said.
It is not difficult to find out that in addition to the numerical data, the machines will also generate log files based on words and numbers, and that these log files have not been used appropriately on the application of fault detection classification (FDC) systems, which uses mainly numerical data, Lu indicated.
Log files are a type of text data which contain some subtle signs that have not been detected before, Lu stated.
An enterprise normally uses server products, including AP servers, database servers, and front-end integrated servers plus third-party watchdog apps to collect data and conduct real-time monitoring during production process, and therefore a sudden accident from this link will interrupt the monitoring process on hundreds of its production machines, which could then to lead to suspension of all the operations of a factory.
But through the examination of log files, it usually can show traits of minor signs such as abnormal connection between AP servers and database servers, which eventually lead to the malfunction of production equipment.
Using SAS's stream analysis technology, which combines the use of natural language processing, to conduct continuous real-time, low-latency analysis on log files, allows an enterprise to stay aware in advance for days or hours before a possible major malfunction, Lu said.
SAS' event stream processing platform supports multiple algorithms and machine learning technologies, utilizing AI edge computing to complete its stream analysis, which not only assists the manufacturing industry for equipment failure warning, but can also be used for defective classification and product quality monitoring and analysis, Lu said.
Bruce Lu, senior consultant, Greater China Pre-sales Support Department, SAS Institute Taiwan
Photo: Elisha Hung, Digitimes, October 2018