top of page
beequip-DrUc34KJJ7M-unsplash.jpg

MASTER THESIS PROJECT

Degree Project in collaboration with Volvo CE, Thesis Final grade: Very Good (B).

beequip-DrUc34KJJ7M-unsplash.jpg

project OVERVIEW

Data-Driven Wheel Loader Operator Behavior Visualization
 

Responsibility: User Research, Concept Design, Visualization
Keywords: Data-driven concept design, Data-visualization, User research, Decision-making, Digital twin
Duration: 6 months
Sponsors: Volvo Construction Equipment

abstract

To realize key business capabilities and secure long-term growth, Volvo CE set out to define a vision for digital transformation. The latest trends in AI-powered by smart electronics open up endless opportunities to help Volvo's operators use Wheel Loaders – Construction machines to increase productivity. These concepts can also be predicted and evaluated using simulation tools. 

To ensure operators are working in a way that delivers optimum fuel efficiency and productivity to achieve optimum results on-site, the company aspires to create visual tools to keep track of operator behavior in the operator environment. Monitor operator behavior with key indicators then visualized to inform how this affects important results for the customers and for Volvo CE. The audience is operators themselves, and internal staff like UX engineers and Product owners.

Thus, this work aims to explore appropriate data visualization techniques under the Data-driven concept design (DDCD) framework. The result is to help Volvo CE, primarily via data visualization, keep track of operator behaviors, and how these affect wheel loader productivity and energy efficiency data on different levels and in a wider context. To carry out, A series of DDCD cases for the improvement of wheel loader operator behaviors are researched and designed, to present data in a clear and concise visual way for both internal audience and operator training.

research questions

1)What methods can be used for the data visualization concept designs? 

2)What features should a visualization of behavior data have to aid internal audiences (UX engineer, Product owner, etc.) for operator behaviors and related results in understanding and decision-making?

3)What features should data visualization have to aid operators in training/feedback/coaching purposes?

case study

The thesis follows the Data-driven concept design and data visualization principles since this work is data-based and prototype-focused. In each stage of the design process, the operator and their demands can be the focus of the DDCD approaches, and gain insights via qualitative data. This method inspired a better user experience and effective data visualization. It helps me to base my design decisions on solid information regarding the operator's and stakeholder's actions, attitudes, needs, etc. The digital twin concepts can be another important reference, to see if validating all the behavioral data in collection and aggregation logics mapped the results at different levels.
The thesis's scope is limited to Wheel Loader in Quarries and Aggregates in a few selected work steps and applications. Under company confidential issues and the industrial supervisor's request, projection onto convex sets via quantitative data is not required.
1.png

💎Data-driven concept design structure greatly helps me in the implementation phase, especially in evaluating design concepts and improving the design process. Since my work gains insights from both data and experience perspectives, this process helps me plan the design concepts and evaluate the design outcomes. To achieve successful data visualization, I need to think about the 💎Data visualization strategies from General principles, Human visual systems, and methods. So I greatly refer to these theoretical foundations to identify the principles that can be used in that. In addition, my industrial supervisor suggest I look at the 💎Digital twin concept for my work, which can be used for online operational monitoring, prediction, optimization, and operator training. So in the visualization design, I also consider the physical system in the real world, the virtual representation in the digital world, and the connection between the two.

3.png
2.png

Pre-study

Then I plan my methodology for this research. To seek and understand the case study context, scenario, and previous solutions proposed by the internal audience, I accessed the study and logs and operator handbook provided by the company. Understand how the wheel loader works and take a driving test onsite. In addition, I read a previous internal study report, to get myself familiar with the short-cycle loading task, especially the classic V or Y pattern. In my work, this is the main research objective. My design is based on this pattern and model. This is the defined wheel loader working system in the real world, then I am going to design the virtual representation in the digital world and define the connection.

Short-loading-cycle-phases-and-extension-from-V-to-Y-cycle-1_W640.jpeg
Pre_study_2.jpeg
Pre_study_5.jpeg
Pre_study_3.jpeg
Pre_study_4.jpeg
Pre_study_1.jpeg

user research

To better understand the internal audience’s expectations and requirements for the visualization, I conducted user research for the two target user groups: the internal audience and operators. The user research is divided into three phases: workshop, interview, and personas. The workshop intended to identify key metrics that the internal audience pays attention to. In the internal workshop, I discussed with three UX engineers and one product owner, to see what Key performance indicators they expected from the visualization. After the internal workshop, I went to the aggregates sites to interview operators. I interviewed two site managers and one experienced operator in Eskilstuna. 

User research

•Workshop

•Interview

•Persona creation

Target user groups

•Internal audience (UX engineers and product owners, etc.)

•Operators (and site managers)

6.png

As the user research result, the scenarios and phases are divided into bucket filling, traveling, and bucket emptying. The general classification of data is State, Physiology, Interaction, Environment, and Phase assessment. The internal audience (focus on UX engineers) and operator personas are created based on results.

Data selection for each phase table.jpg
8.png
9.png

prototype

The prototype was developed using continuous feedback from the stakeholders at Volvo CE. It consists of the preparation and analysis of internal user research data and operator interview data to produce relevant user knowledge for data-driven visualization. It also follows the process of Human-centered design for interactive systems and refers to the digital twin concept. I transform the classic short-cycle loading into digital form and define data as the connection between the two worlds.

The prototype included two visualization deliverables, a data-informed dashboard for the internal audience (mainly UX engineers, and product owners) for decision-making, and a data-informed dashboard for operator coaching/training, to distinguish it from the first one, it was called the digital tool. 

 
 
Interaction techniques selection table.png

Based on the literature review and user research results, I benchmarked all data against the common data visualization techniques, then made the design decision and sketched them in the low-fidelity prototypes. It included the shapes of elements, basic visual hierarchy, and information layout, which were considered key elements of the visualization content. It allows the internal audiences to explore different ideas without too much effort. Feedback and insightful ideas were gathered in the continuous workshops and meetings, they were later circled back and used as inputs for iteration.

 

Low-fidelity prototype - A data-informed dashboard for the Internal audience (UX engineers and product owners, etc.)

10.png

Low-fidelity prototype - A digital tool for Operators (and site managers)

11.png

Next, I applied visualization and interaction techniques to the high-fidelity prototype. 

Here is the data dashboard. The data dashboard is designed for the internal audience like the engineers, to help with decision-making.

 

This is the digital tool designed for the operators, it is like the previous data dashboard but with more focus on the behavioral comparison.

In this digital tool, operators can view their behavioral data in a period and compare it with other operators. Data from two individuals will be presented on the board, and an auto-generated insight will show on the panel also. Due to the GDPR, all the information is depersonalized. So that the operators can focus on their behavior data and insights themselves and learn lessons to do better work next time.

 

evaluation

In the evaluation, participants show great interest in the dashboard and digital tool. They provided insightful feedback on my work, I received 4 questionnaire answers from the experts after the evaluation meeting session, their roles vary from UX engineer, Software engineer, and data analyst. 

图片 1.png

Over 75% of participants have prior experience in wheel loader driving and gave positive feedback to the visualization overall designs and thought they can get insights and information from the data presented. All of them are familiar with the visualization techniques used in the dashboard and digital tool, but most of them present a neutral attitude to the question regarding the visualization's intuitive and faster decision-making. In the visualization techniques selection evaluation, the most effective type is the pie chart, bar chart, table, and line chart. The digital tool for operator training received more positive feedback for its intuitive way to compare performance among individuals.

Internal evaluation results .png

discussion

Then I discuss whether the results are as expected. From the methods perspective, the initial tool was Power Bi but then was replaced by Figma. Because my work was mainly focused on the conceptual design, not the development. And we didn't receive the qualitative data from the company due to credential issues. In the evaluation part, I initially wanted to set up the live evaluation experiment combined with a 30 mins 1-to-1 expert interview. But due to the time limit, we didn’t find time to conduct the 1-to-1 interview in the final.

The discussion of the result, from the evaluation, users have different ideas on the time unit definitions. The internal audiences wanted to check data from various time levels. They focused on both cycles, daily, and weekly levels. But in the design, I must make choice in time unit definition design. The made designs was can be accessed through the daily level, weekly level, and cycle level. And from the design part, the visualization methods can be improved to get closer to operators' specific behavior in short cycle loading tasks.

Thus, in the future, mechanical and technical perspectives can be look into, adding more key factors, like onboarding time and idle time and speed heatmap, season/time, machine and machine type, and adding the personal filter. If the conclusions drawn from the qualitative metrics produced from the user research want to improve, more stakeholders must be involved. The definition of insights was a bit arbitrary during the user research. Supported tools like Mixpanel can be used for user engagement measurement and behavioral efficacy, to be a significant reference in future work on data visualization.

conclusion

In conclusion, answering all research questions. The research conducted in this thesis shows that DDCD is the essential method in the visualization concept design process. Also, the theory of data visualization greatly contributes to the visualization technique selection. The digital twin's concept help construct an architecture that connects the physical wheel loader and the digital visualizations.

The dashboard for the internal audience included seven parts: site and time selection, weekly overview window, phase selection, cycle thread trace, insight window, data presentation, and toolbox. These features have the capabilities to aid internal audiences with operator behaviors and related results in understanding and decision-making. The visualization for operators also known as the digital tool consists of seven parts: site and time selection, opponent selection, phase selection, cycle thread trace, the external data window, individual comparison section, and insights block.

According to the evaluation, most of the evaluation participants gave positive feedback on the designs, the internal audience affirmed the value of the visualization. Over 75% of participants gave positive feedback on the digital tool for operator training.

reflection

I also reflected on my work and think about sustainability issues. By using the dashboard and digital tool, operators' behavior can not only be measured and analyzed but also improved. This gives Volvo CE the opportunity to establish effective fuel efficiency drive practices while creating more profitable and sustainable transportation. Through user research and focuses on user behaviors, the final goal is to increase energy efficiency and reduce CO2 emission, strive to build and be part of an environmentally and socially sustainable value chain, and answer the call of the 2030 Agenda for Sustainable Development by the United Nations.

The major limitation of the present study is that this work only adapted to selected work steps in wheel loaders in quarries and aggregates. Since the lack of a quantitative sample and mathematical model from the host company, the data visualization doesn't project onto convex sets. Moreover, this data visualization design process is restricted to conceptual design in the internal environment, not including the subsequent business product development with data analysis.

bottom of page