Analytics have become a big part of manufacturing operations. For decades, plants have generated and collected data, but modern technologies with advanced analytics, machine learning (ML) and other artificial intelligence (AI) algorithms are taking how we use this data to new levels. The intelligence gleaned from the data enables continuous improvement of processes and products, helps builds quality into processes, and provides a feed to predictive algorithms that can anticipate events before they happen.
Advanced analytics cover descriptive, diagnostic, predictive and prescriptive analysis.
This means we can understand what has happened (descriptive), why it happened (diagnostic), what will happen (predictive) and what actions should be taken (prescriptive). Armed with these four ‘heroes’ of information, steps can be taken to greatly improve plant efficiency, avoid processing errors, reduce waste and re-work, and enhance quality.
Analytics can be designed to do pretty much anything for a manufacturing operation. They can be used to identify patterns, correlate cause and effect and model complex processes. With enough data, learning algorithms can be used to approximate any function on the shop floor. Modern manufacturing execution systems (MES) collect, correlate, contextualize and analyze data throughout the supply chain. Representing a single source of truth, powerful analytical capabilities with these systems mean they can quickly analyze data to find new ways of driving efficiency in the production process. The faster the meaning of data is understood, the more agile a process becomes – and each second rescued can save money from the bottom line.
Exploring the Types of Analytics
Descriptive analysis is generally the preliminary stage of analytics. It aggregates and summarizes data to give insight into what has happened through the manufacturing process. It uses historical data to draw comparisons and understanding about changes that have occurred. Output from this analysis can be presented in reports and dashboards to give a clear picture of trends and anomalies.
Diagnostic analysis takes the information aggregated by the descriptive analytics and looks to why any changes occurred. A form of root cause analysis, it uses advanced analytical algorithms to look for correlations and provide deeper insight into why something is changing. Diagnostic analytics takes different historical data sets and looks for trends or anomalies to determine relationships and, therefore, better understanding of cause.
A use case for Diagnostic Analysis is with Smart Modular Technologies in Brazil. SMART manufactures sophisticated memory products including DRAM, and LPDRAM, eEMC, eMMC, and microSD, and as the production volume grows, managing flows in their internally developed WIP tracking system and recording data in Excel did not allow for close yield monitoring or data analysis to prevent defects. It was replaced by complete automated data collection with standardized production flows. As a result, SMART achieved full WIP traceability complemented by quality and SPC, recipes and equipment selection and tight control of time windows between processes.
Predictive analysis offers exciting opportunities to increase productivity. The most obvious and widely used example is predictive maintenance, whereby data collected from machine condition monitoring sensors over a long period of time is merged with previous maintenance activity. This enables the system to recognize markers in real-time performance to determine when maintenance will be needed, rather than simply following a set maintenance schedule. In this way, maintenance will only be carried out when it is required, unnecessary maintenance activities are removed, and production schedules can be fitted around maintenance schedules to maximize overall shop floor productivity.
Predictive analytics opens up many predictive modelling scenarios. This gives manufacturers the opportunity to take action to mitigate events that will impact production efficiency. In the wider supply chain, these predictive models can help improve inventory and logistics management to make the overall operation more efficient. They can be used to improve yield. Correlating and analyzing data, production engineers can determine whether the use of a particular machine or parameter have a positive or negative effect on yield or performance. In this way, workflows and processes can be adjusted for greater optimization and better production results.
Prescriptive analysis is the last of our four analytical heroes. Having determined what happened, why it happened and what is likely to happen; prescriptive analysis works to find the best course of action to optimize manufacturing processes. It takes the huge volumes of data from the smart factory, which would take many hours, days, or even weeks to assess manually, and presents the best tactical choices. Using ML and AI, prescriptive analysis can become smarter over time and has the potential to identify new rules or procedures to be followed.
Data is at the heart of a smart factory and, as volumes of data grow, they can be used to feed learning and predictive algorithms to increase levels of efficiency and automation across all shop floor operations. With the right data inputs, analytical algorithms can be written to find information that can help with process performance, product quality and overall business strategy.
The modern MES provides a single repository for data and a way to quickly analyze and present results back in a familiar form. These systems incorporate AI technology and are capable of all types of advanced analytics to give clearer pictures of what has happened, why it happened, what will happen and what should be done. With the answers to these questions, we leap forward in the battle for efficiency.