Plant Data Visualization Best Practices

/

Providing access to plant data at the right level and time is critical to effective operations and key to achieving the organizations targets. Decision makers need a clear understanding of the current situation and overall performance. Understanding trends and relationships enable experts to analyze and optimize the process and forecast the future. Operators are required to quickly spot issues in real time to take corrective actions. A lot of effort and investment goes into building the infrastructure for gathering, storing, and processing of plant data. Effective delivery of information to users often requires the use of specific Data Visualization techniques.

Figure 1: Role of data visualization in decision making

Data Visualization can be defined as “a collection of methods that use visual representations to explore, make sense of, and communicate quantitative data” [1]. The main goal of visualization is to improve the users understanding, and therefore requires knowledge of how users perceive data and interact with it. Studies in human perception provide guidelines on what users can observe and remember in a short time span. Likewise, domain knowledge is necessary to identify the most important pieces of information to the users. Aesthetics and responsiveness of the design help increase user engagement, while taking care not to sacrifice the quality of data presented.

The recent advancements in technology have lowered the deployment cost of visualization solutions. High resolution displays enable the rendering of sharper graphics with readable information on both large monitors and mobile devices. Standardization of web graphics in HTML5 enable rapid deployment of dashboards and interactive graphics on both desktop and mobile phones. Faster processing speed allows for a more responsive user experience while interacting with visualization. Additionally, the recent rise of cloud analytics solutions enables quick integration of business intelligence capabilities to user data. User expectations of visualization systems have risen accordingly.

Plant Information Dashboards

Dashboards have become a popular way of visualizing plant information. Information in dashboards are often represented as Key Performance Indicators (KPIs) that are calculated by processing data from multiple sources. Dashboards may be deployed as part of collaboration portals that enable different departments and functions to exchange data and execute business workflows. Multiple dashboards may be deployed to display different levels of information according to user roles and areas of operation. While dashboards are mostly used for real-time monitoring, detailed reports and drilldown views are often provided for further exploration and analysis.

Dashboards enable users to quickly become aware of the current situation of the plant and are often displayed on dedicated monitors or large displays that show the progress as the day goes on. Visualizations are used to add context to the displayed data. For example, gauges highlight the desirable targets of the plant and how close they are to be achieved, comparison charts show the contribution of each unit to the overall outcomes, and trends enable an understanding of behavior over a specified period of time.

Applying data visualization techniques is essential for effective dashboard design. The main goal is to focus on most critical information while reducing the noise generated from displaying excessive data or non-data visuals. Efficient use of the limited available space is required, especially on smaller display mediums. Providing visual cues that help the user quickly identify patterns or interact with data is essential. Furthermore, flexibility and continuous improvement of design as the user requirements evolve ensures retention of user engagement.

 

Figure 2: an example of a modern plant dashboard

 

In our experience in deploying dashboard solutions, we find the below guidelines very helpful in delivering effective solutions and meeting user expectations:

  1. Engage End Users Early
  2. Follow Proven Design Principles
  3. Focus on the Critical Data
  4. Choose the Right Visualization

Engage Users Early

It might occur that end users start interacting with the dashboards at or after project delivery, which may lead to gaps between the designed solution and user expectations. In our experience, following the agile process for development of the dashboards while engaging users with questionnaires, samples and prototypes helps closing the gap and reduce the overall time and effort required for delivery. It is recommended to involve domain experts and consultants to help users identify their requirements and arrive at the most effective methods for achieving them.

Figure 3: iterative process of dashboard desig

Here are some examples of questions that help users clarify their requirements:

  1. How frequently is the dashboard updated? (Days, Hours, Minutes, Seconds)
  2. Who is going to use the dashboard? (Management, Operation, Maintenance, Financial, Quality, Safety, etc.)
  3. What information is required to display on the dashboard? And how are they calculated?
  4. What are the actions taken based on the information displayed in the dashboard?
  5. What is the hierarchy used for organizing the dashboard (plant, area, unit, etc.)
  6. What are the targets for values displayed in the dashboard, and how are they obtained? (daily production targets, HSE regulations, etc.)
  7. Where are the dashboards going to be displayed? (Workstation Monitors, Video Walls, Tables, Mobile, etc.)

Follow Proven Design Principles

Time constraints and pressure to complete the desired dashboard functionality may cause fundamental design principles to be overlooked. On the other hand, a tendency to rely on personal taste leads to subjective design choices that might not be well perceived by other users. An experienced designer would incorporate the design principles in their workflow and take user feedback into the process. Techniques such as A/B Testing can be used to measure user response to different choices in design.

Literature and studies on design principles are numerous, interested readers can refer to the Further Readings section for additional references. In particular, we found the following principles to have great effect on improving the quality of dashboards:

  1. Use colors to encode meaning. Random choices of colors may confuse the users. Different Hues can encode different categories of data while color saturation may encode emphasis. However, it is important to avoid over using highly saturated colors or hard to distinguish hues as they lead to eye fatigue or loss of information.
  2. Users are quick to detect patterns instead of remembering sparse data. Use space, orientation, shapes and sizes to form patterns. Use borders to group relevant data together. Avoid over-crowded plots that show very high frequency data or large number of trends, instead, filter or zoom in the data to the correct level.
  3. Maximize the data ratio in visualizations. Try to remove extra graphical elements that add little or no data. Give priority to flat, solid color visualizations over 3D or unnecessary gradients. Remove unnecessary lines and keep the background colors consistent.
  4. Aim for simpler and easy to understand design over complex and hard to change graphics. Complex, multidimensional visualizations may be broken into simpler ones using techniques such as Small Multiples for better results.
  5. Provide clear context to visualizations in terms of time, area, scale and reference to avoid confusion.

Figure 4: Example of using small multiples of the same chart to compare production/consumption across different plants. By using the same range in all charts, it is easier to compare quantities across the plants.

Focus on the Critical Data

Dashboards are ideally designed to fit the screen. Scrolling the dashboard to view rest of it is not desirable as it leads to loss of important information. The limited available space requires prioritizing which information to be displayed. The following techniques help focusing the content of the dashboard:

  1. Well defined KPIs provide a great source of brief and concise information.
  2. Summaries and roll-ups of aggregated data would be more suitable to show on a dashboard, while further details may be explored in drill-down analysis.
  3. Using rounded up values, percentages or multiples (e.g., thousands, millions) instead of high precision data might save some space without general loss of quality.
  4. Short lists of Top/Bottom performers help focus the user attention.
  5. The use of space-efficient visualizations such as Bullet Graph and Sparkline helps conveying the information without loss of space.

Choose the Right Visualization

Visualizations are created for different purposes and can convey different aspects of the provided information. The choice of visualization should be driven by the objectives that we would like to achieve. Moreover, many visualizations can be fine-tuned using several parameters such as scale, range, and number of categories, etc. Poor choice of parameters might lead to loss of important information that the user would like to observe.

Gauges are commonly used to visualize KPIs, they are often modeled after physical gauges that are familiar to the plant operator. Gauges use combination of indicator and color to mark the current status against a certain range. Target values are usually marked on the scale while alarm regions are marked using standard yellow and red colors. While the physical resemblance makes the gauge familiar to the users, more abstract designs may be used to visualize KPIs in smaller space.

 

Figure 5: Gauges are the de-facto visualizations for KPIs

 

While gauges represent the current value, other indicators can be used to show the KPI trend. Arrows can display whether the value is increasing, decreasing, or generally constant. Sparklines can provide a more informative trend over a set period of time.

 

Figure 6: Sparklines add historical context to current KPI value

 

Line plots are used to display trends over a period of time. When designing the trends, it is important to select proper scales and labels for the axes. Horizontal lines can be used to plot thresholds for data while vertical lines can be used to plot times for important events or differentiate between actual data and future predictions.

Bar charts are used to compare quantities of categorized data. Similar to trends, labels and scales should be defined properly. Categories on a bar chart can be sorted according to their values or based on a user defined order. Quantities in a bar chart can be stacked to show further details in values. Bar charts are efficient in conveying differences in quantities and are generally more preferred to other plots such as pie charts [2].

Scatter plots are used to compare two or more variables in a data set in a two dimensional plot. Two variables can be plotted on the X-Y axes while additional variables can be encoded using color, size and shape of dots. Depending on the size of data, scatter plots might become very dense and hard to interpret. Further processing of data can produce Heat maps with color coded regions in a two dimensional matrix for easier detection of correlations.

Other visualizations that can be used in dashboards include box plots and histograms for statistical distributions and tree maps for hierarchal data.  

The Way Forward

The KPI team at Yokogawa consists of close-knit performance analysts, system professionals and domain experts that are ready to assist in conceptualizing and implementing the appropriate KPI based system.  KBC and Yokogawa experts can bring a new era of prosperity to our client’s enterprise, ensuring maximum value addition to the end-user of the KPI project by way of performance boost and effective decision making. To learn more about Yokogawa's Performance Dashboard Solution, click here.

Further Readings

1. Few, Stephen. Information dashboard design: displaying data for at-a-glance monitoring. Burlingame, CA: Analytics Press, 2013.

2. Ware, Colin. Information visualization: perception for design. Waltham, MA: Morgan Kaufmann, 2013.