Keywords

1 Smart Service Operation

1.1 Augmented-Reality

Augmented Reality (AR) is a technical option very much hyped for industrial use. AR is the form of Mixed Reality, where virtual content as computer generated virtual objects or environments is blended into the real world [1, 2]. The user therefore stays connected to the real world which is mandatory for service operation. For industrial service operation AR allows an interactive way of presenting i.e. maintenance information along the service procedure, typically under the use of smartphone, tablet or Head-Mounted-Displays (HMD). It is a new way for equipment manufacturers to deliver maintenance instructions for maintenance specialists and/or equipment users [3]. AR seems a promising tool to ensure a proper, quick and safe maintenance fulfillment and may enrich a service technicians’ workplace with modern technique.

However fundamental research on the positive or negative, on the enabling or stressing effects of AR use within service operation is missing. Research of AR-use in typical manufacturing tasks as i.e. show positive effects of AR use [4, 5] But does AR ease the work circumstances and therefore supports a safe and healthy workplace and increases performance for service technicians? To find about this, the authors executed a first experimental study comparing differences in workload, performance and operational safety when using paper-based work instructions vs. AR-based work instructions within a maintenance scenario.

1.2 Industrial Service Operation

Industrial services must assure a safe, economic and compliant operation of industrial assets as i.e. paper mills, power plants or trains. As there is an increase of automation and human-computer-interaction in the asset operation, there are comparable changes within the service operation of these assets. Smart technologies as i.e. Head-Mounted- Displays, exoskeletons or smart gloves are promising tools to lower work strains [6]. However, the effects of automation on workplace, OSH and work motivation are even more relevant in the industrial service domain. Due to a low level of repetitive tasks and heterogeneous, highly customized service operation, the human worker will not be completely replaced. Therefore, studying the effects of automation on industrial service technicians is of long-term interest.

Work instructions play an important role for a safe, successful service operation because

  • As the variety of problems due to an increase of physical assets (i.e. manufacturing machines) and asset properties (i.e. programming language, functionality) increases the number of customer-driven service tasks for which the technicians has to look for instructions increases.

  • Another phenomenon in some regions and industrial service sectors is the employment of low-qualified workers for whom work-instructions are the only way to fulfil the service task.

  • Industrial service technicians often have to work in a more hazardous work environment. Safety instructions coming along the work steps are sometimes numerous, must be confirmed by the technician and filed by the employer.

AR offers a smart information providence supporting the service technicians. AR-working instructions give an automated step-by-step guidance depending on the service object and previous working steps. It can be understood as a solution with a low level of automation (LOA) [7, 8]. Through the use of object recognition the technician gets detailed, accurately fitting information helping to fulfil the work routine which the technician confirms and executes. Following Parasuraman and colleagues’ (2000), the third LOA (out of 10) is therefore implemented [7].

In comparison, within paper-based work instructions the technician must decide on himself which information is relevant in each working step. There is no situation-adapted automated information providence.

AR-use is a form of human-computer-interaction, where a change in workplace, its procedures and safety issues are to be expected.

1.3 Workload and OSH

Workload is the entirety of all requirements and external conditions at a workplace, which could influence a person physically and/ or psychologically [9]. Workload defines the strain of a task. Both can be understood as synonyms. Within the NASA-TLX the identified components of workload are mental, physical, temporal demand, effort, performance and frustration level [10]. Especially mental workload is of interest in industrial service operation. Heterogeneous work conditions defined trough differences in the assets, the customer requirements and an increase in asset complexity causes mental strain. Temporal demand is as well typical for service operation. In addition, service technicians’ workplaces are typically noisier and dustier [11].

Mental workload is seen as the sum of all external parameters influencing the employee mentally [12] and is described as “the degree or percentage of the operator’s information processing capacity which is expended in meeting system demands” [13]. Operational features causing mental workload are amongst others

  • Lack or overflow of information

  • Difficulties in information acquisition

  • Conflicting task assignment

  • Work discontinuity [14].

Mental workload has a direct input on the task output (Fig. 1.) This is defined by the work quality in relation to the time needed for fulfillment [15]. Service quality is defined by a reliability, assurance, tangibles, empathy and responsiveness [16] and strongly depends on the manpower’s capabilities [17]. Therefore service quality incorporates operational safety and health (OSH) as a crucial factor for reliable service operation. OSH can be achieved through (a) good ergonomics, (b) active or passive safety features and (c) personnel qualification and organization. Working instructions have an influence on the personnel qualification and should reduce the effect of the outlined operational features causing mental workload.

Fig. 1.
figure 1

Output and mental workload [18, 19]

Full size image

1.4 Hypothesis

If AR has the positive effects on mental load factors and gives automated information on-demand, the following hypothesis are made

H1:

AR will cause a significant reduction in the amount of time to complete a service task compared to paper-based work instructions.

H2:

The number of errors done by the technician is reduced by using AR compared to paper-based work instructions.

H3:

Subjective workload measures indicate a decrease in mental workload for the technicians

H4:

The positive effects are appreciated more by younger workers.

H5:

Safety instructions are followed better by the use of AR.

2 Study

This study took place January 2020. In total, 20 subjects took part - all of them non-service-technicians in real life. The subjects were grouped in two – paper and AR. The two groups showed the following age and gender characteristics (Fig. 2). Gender and age were selected corresponding to the characteristics in real world, where female technicians are rare [11].

Fig. 2.
figure 2

Subject groups characteristics

Full size image

The survey included the NASA-TLX by Hart et al. [10], time, quality and safety observations, and demographic variables. The subjects were asked to rate all six demand categories (mental, physical, temporal, effort, performance and frustration) using a 100-point scale. The subjects were then asked to perform pair-wise comparisons to give an individual feedback on which of the load categories of the TLX are more important. The rates of each category and the result of the pair-wise comparisons lead to a mean weighted workload score – the overall task load index.

2.1 Experimental Setting

20 subjects were asked to repair a chain saw whose starter pull cord has been broken (Fig. 3). They were asked to fulfill the task as quick as possible. None of the subjects had ever done this repair before. 10 test persons were asked to fulfil the repair scenario under the help of a printed work-instruction, 10 persons could use the help of an AR-work-instruction installed on the Microsoft Hololens®.

Fig. 3.
figure 3

Experimental setting

Full size image

7 out of 10 subjects who used the AR-implementation had no experience with AR at all, 3 had low experience. All 10 got a brief introduction in how to use the Hololens® and were asked to go through the gesture control app preinstalled on the Hololens® by Microsoft.

2.2 Findings

Time of Completion.

Time of completion is a crucial factor evaluating service output. In general, while service is done, the industrial asset is not available. Unavailability means financial loss due to reduced earnings. Remarkable, the average time of completion using AR-based work instructions is about 35% less compared to subjects using paper-based work instructions (Fig. 4).

Fig. 4.
figure 4

Effect of instructional medium on time of completion.

Full size image

However, ANOVAs (using an alpha level of 0, 05) comparing the effect of instructional medium on time of completion showed no statistical significance, F(1,18) = 3,265, p = 0,199. The age has as well no significant influence on the time of completion.

In addition, not only time of completion is a lot less using AR-based instructions, also the subjective evaluation on how well the repair process was completed shows differences even though all subjects completed the maintenance task successfully. 80% of the subjects using AR-based instructions rated their work result as good or very-good, only 50% of the participants using the medium paper rated their work result as good, none as very good (Table 1).

Table 1. Subjective feedback on how well task was fulfilled
Full size table

Workload.

There is a noticeable difference in mental workload between the two mediums. Descriptive statistics show an average mental workload of 56 when using paper-based instructions compared to 27, 5 (Fig. 5). A one-way ANOVA shows a statistically significant effect of instructional medium on mental workload, F(1,18) = 13,048, p = 0,571. Due to the interference of computer-generated overlaid, animated objects, it is of less cognitive demand to find out about what is meant by the working steps and which parts and tools are needed.

Fig. 5.
figure 5

Effect of instructional medium on mental workload

Full size image

A one-way ANOVA on the effect of age on mental workload showed no statistical significance. However, Post hoc comparisons showed that there is a statistically significant effect between ‘age group < 35’ and ‘age group > 50’ (p = 0,085).

For the overall TLX there is still a difference in favor for the AR-based work instructions (Fig. 6), however this difference is not statistically significant, F(1,18) = 3,884, p = 0,875. This is mainly because besides the mental workload there is only a noticeable, but not statistically significant difference in temporal demand (Fig. 7). The other four demand categories are very much independent of the used medium.

Fig. 6.
figure 6

Effect of instructional medium on overall TLX

Full size image
Fig. 7.
figure 7

Effect of instructional medium on temporal demand

Full size image

Operational Safety.

The service scenario was successfully completed by all 20 subjects. However, only during the use of the paper-based instruction most subjects needed help by the instructors (8 out of 10 subjects) as they were unsure in certain steps. As with the chain saw a realistic object was used, the instructors interfered in cases where there was a potential risk to damage the chain saw. In AR computer-generated instructions are overlaid to the real object. This makes it much easier to figure out correct positions, orientations, parts and tools (compare mental workload). None of the subjects got a technical instruction to the chain saw. But the noticeable difference in queries show that the level of safety relevant knowledge is higher after the same time when instructed by AR compared to paper-based work instructed. The findings underline previous research where it is stated that AR applications “are cheaper and more efficient ways to enhance human safety” [20]. AR allows a better safety relevant qualification in shorter period of time for untrained employees. The participants using AR-based work instructions declared the work-instruction much more helpful than the participants who had to use paper-based work instructions. A one-way ANOVA was conducted on the effect of instructional medium on how helpful the instruction was. The effect is statistically significant, F (1,18) = 6,4, p = 0,913.

Safety Devices.

Looking at the usage of protective clothing as passive OSH measurements, no differences could be seen. In both groups 60% of all subjects put on the required safety devices gloves and glasses. AR-use seems to have no influence - neither positive nor negative - on the use of safety devices. This result is unexpected as the AR-work instruction did not only include safety information like in the paper-based work instruction, but also asked the subjects to confirm the information by gesture or voice command.

3 Conclusions and Outlook

AR-based work instruction can be understood as an enabling technology in service operation. Service output defined as the work quality in relation to the time needed for fulfillment is enhanced. Even though paper-based work instructions are well established and standardized, AR offers a big advantage for untrained staff as task-relevant information is overlaid to the physical object. Uncertainty in what and how to do each service step does not arise. Time of completion and mental workload are reduced by AR-based assistance. The self-assessment regarding the task output, a potential motivator, is more positive. However, AR implemented even in the MS Hololens®, as one of the most advanced solution available right now, still has some obstacles hindering the technician in the tool use: (a) the field of view of MS Hololens® 1 is very small – the user finds it difficult to see all animations and the control panel in the field of view at the same time, (b) the weight of the HMD (579 grams) is heavy for long-term use - some subjects felt pain after a short period of time of approximately 20 min.

AR can be a good alternative to the well-established but rather laborious printed handbooks. Looking at the hypothesis, following conclusions can be made from this study:

  • H1 may be neglected. There is evidence that AR causes a significant reduction in the amount of time to complete a service task compared to paper-based work instructions. However, the effect is not statistically significant. Further research and a bigger sample should be done for verification.

  • H2 may be neglected. No subject did an error, all completed the repair process successfully. However, the number of queries is significantly higher in the paper-based instructed work. Without the interference by the instructors the number of errors is expected to be much higher in the paper-instructed service operation. For further studies adoption in the scenario should be made.

  • H3 is confirmed. There is a significant decrease in mental workload for the technicians using AR-based work instructions.

  • H4 is confirmed. The positive effects are appreciated more by younger workers. The participants younger than 35 had a significant lower mental workload then the ones older than 50 years old.

  • H5 must be neglected. Safety instructions are not followed better by the use of AR. Even the necessity to confirm safety instructions didn’t lead to a higher degree of safety device use. In addition, as all subjects completed the work task successfully, there is no direct effect on OSH by an observed decrease in the number of errors. However, if instruction are understood as a safety tool, AR is much more appreciated then paper-based work instructions.

This study has only looked at the changes from printed to AR-based work instructions. For further research, video instructions as a reasonable, wide-spread option should be included in an evaluation on how smart information use within industrial service operation causes positive and/or negative effects in the technician’s workplace.

Further did this study only focus on short-term effects. Long-term effects also with regard to the customer who normally is somehow involved in the service operation has been left aside for this study.

In addition, further research should address possible influences on work motivation coming along with smart information use in service operation. As service performance increases, there might be a positive long-term effect on work motivation leading to many more effects than the ones looked at within this paper.

Notes.

As the study is based on a realistic maintenance scenario, the results have to be understood as a comparison between (A) the standardized, well-established medium paper-based work instruction, which was provided by the manufacturer and is so far used for the chain saw maintenance and (B) a prototypic, non-standardized AR-based work instruction whose animated information sequences are based on the paper work instruction.