Apr-2023

Importance of effective and efficient data analysis and visualisation

Kit-based ‘out-of-the-box’ dashboard solutions address many industry challenges, both in making better and faster decisions and simplifying and accelerating solution delivery.

Philippa Hayward
KBC (A Yokogawa Company)

Article Summary

Throughout history, scientists have observed our world and its processes primarily out of natural curiosity and to improve things. The ability to visually see performance and graphically analyse it is, today, a powerful method to drive performance of an asset and reduce carbon emissions and energy consumption.

Regulatory compliance, including environmental and health and safety regulations, mandates certain aspects of performance analysis and reporting. Currently, for example, there is an increasing emphasis on monitoring a company’s direct and indirect carbon emissions (operated facilities and supply chain). Thus, analysis and reporting are crucial in complex process industries, such as refining, to ensure performance stays on target. This information enables operators and engineers to take pre-emptive action to prevent process performance from straying from the expected operating window. The performance of equipment, catalysts, instruments, and humans can all contribute to negative deviations from expected performance.

Many companies have already taken steps to make data widely available for analysis. Although digitalisation initiatives are driving automation of analytics and shortening decision times, integrating these analytics into key work processes remains a challenge.

Business challenges
Business and engineering teams face many common challenges when analysing their processes to keep them on target and improve them. Understanding these challenges is essential to designing effective data analysis and visualisation tools.

Productivity
Accessing data and analysis tools from disparate systems can cause highly skilled staff to waste time on routine data collection and analysis tasks. Instead, this time could be better spent on higher-value tasks. Also, handover efficiency is reduced if engineers have to learn a new set of analytics tools and/or methodology for each new unit.

Missed or slow opportunity identification
A lack of a systematic approach to data analysis can lead to slower decision-making:
• Business teams may disagree on which key performance indicators (KPIs) to use, their relative importance, and how to calculate them. This can result in unnecessary rework, rechecking data, and additional discussions before decisions can be made to improve performance.
• Inefficient and ineffective dissemination of insights can impede the efficiency with which decision-makers act.
• Tools tailored to the process or process engineer may become difficult to use when the ‘expert’ is unavailable due to sickness, vacations, or urgent business. In such circumstances, only safety-critical work may be performed, leaving non-critical but otherwise significant untapped opportunities.

Knowledge transfer
An ageing workforce and frequent staff rotations can make process expertise a highly valuable yet scarce resource. However, teams must have appropriate process insights to make the best business decisions. Incorporating best practices and encapsulating process knowledge into analytics tools could mitigate the problem.

Developing a solution
In today’s industrial environment, monitoring and analysing manufacturing processes is no longer a question of whether to do so but how to do it best. As data availability becomes less of an issue, data quality and connectivity still pose challenges. Solutions range from the well-loved but highly problematic engineering tool of first resort – the Excel spreadsheet – to company-wide information management systems with sophisticated data analytics.
Ideally, the solution should be:

• Fit for purpose: many engineers and IT professionals are tempted to choose the latest, most detailed solution right away because they love technology for technology’s sake. Although designs need futureproofing, conducting a simple cost-benefit analysis should help determine whether all the ‘shiny new toys’ in the box are worth using now, should be saved for later, or are simply inappropriate for this case

• Scalable: attempting to ‘digitalise’ a complete site all at once is challenging for all businesses except the most well-resourced and well-organised. Digitalisation solutions can be delivered stagewise (starting with the unit that will provide the most benefit), leaving provisions for future complexity (starting with simpler data analytics and progressing to a rigorous process simulator-based digital twin and artificial intelligence/machine learning in a subsequent update). Furthermore, the solution can be improved as the rollout progresses.

These solutions need three common elements:
• Reliable, integrated data sources and storage systems (starting with an on-premise plant historian, extending to data lakes and cloud-based storage).
• Analytics aligned with the company’s management needs.
• Insights gleaned from the data disseminated to the actors and decision-makers.

The following discussion assumes that data have been logged and stored efficiently (whether on-premises and/or in the cloud) and are easily accessible. Data silos still exist to a greater or lesser extent in most organisations for historical operational and/or technical reasons. Siloed data can be a major barrier to efficient decision-making. When decision- makers lack access to all the relevant information, or if access is delayed, such as by manual steps to synthesise data, decisions will be suboptimal and/or late. Both situations result in lost opportunities.

In terms of analytics, the first step starts with understanding what should be monitored, analysed, and why. In a typical process unit, this means identifying KPIs and their relative importance. A seemingly limitless supply of data and the computational power to process it are no replacement for engineering and scientific expertise. At the least, data quality should be checked. Poor data offer little value, and inconsequential correlations and relationships in the data should not cause distractions. Engineering skill is also needed to ensure that the scope and complexity of the analytics match the size and nature of the problems.

Data analysis and process monitoring only yield value if they lead to action. Real-time optimisation and automated optimisation solutions minimise the need for human intervention in plant optimisation but are often the most expensive option and unsuitable for many processes. Also, it may be more appropriate (financially and technically) to start with a less complex solution to develop over time. So, to enhance the decision-making process, what are the best ways to present information when considering an advisory solution?