Keywords

1 Introduction

Nowadays, advancing technology in the industry such as automation and mechanization affect the redesign of workplaces, as well as the implementation of new systems in industrial applications (e.g., systems for human–machine cooperation [1], exoskeletons [2], or augmented reality systems [3]). Demographic change forces employers to provide more technical assistance systems to reduce musculoskeletal loads and enable a longer, healthier and safer working life [4, 12]. More than one of three people manipulates heavyweight goods during the workday, 43 percent daily work in tiring, exhausting, or painful postures [4]. As a result, workers are physically burdened and exposed to a risk of developing musculoskeletal disorders (MSD) [4]. Studies of [5] complain about 21.6 billion Euro loss of gross value due to incapacity to work days caused by MSD. Forecasts define a worldwide market volume of up to 5.6 billion dollars in 2025 for the exoskeleton industry, where especially work-assisting devices will grow exponentially [6]. Upper body exoskeletons will take on a primary role for possible future solutions for specific work tasks like lifting heavy parts and for overhead working tasks. Regarding industrial applications, exoskeletons are externally wearable mechanical devices [7] that either empower, facilitate, stabilize, or add movements [8]. Support systems such as exoskeletons are used with the aim of reducing strain on workers without having to make extensive interventions in the work process flow [8]. An ergonomic design of the work process can reduce the development of musculoskeletal disorders but can also become an economic challenge in the case of significant process and product changes [9]. In practice it is difficult and costly to prove the exact effectiveness of exoskeletons for supporting specific work activities. Laboratory and field studies are conducted for this purpose, but they require a great amount of time and expense in product development [10]. Currently, there are no exoskeletons on the market that can be manufactured according to variable parameters and specified boundary conditions. Users cannot find a suitable system that satisfactorily addresses their individual requirements - for example, movements are restricted, kinematic structures do not match movement patterns or interfaces are uncomfortable [10]. The process for a user- and task-specific new design of exoskeletons is influenced by various aspects. The technically complex replication of joints (e.g., the shoulder joint with several degrees of freedom) is always a challenge in system development, in order to ensure that there are no movement restrictions for the user. There is a demand how to validate and optimize exoskeletons for the applied task in respect to movements and loads. End-users of exoskeletons vary in population characteristics such as anthropometry, muscle strength, body mass, manner of executing movements, and each application scenario varies concerning movement and load-specific boundary conditions. Digital human models comprising the human as well as the exoskeleton in a single biomechanical system offer the chance to consider all these aspects in parallel [6].

2 Evaluation of the Biomechanical Requirements for the Digital Twin

The detailed evaluation of the workplace and the working environment as well the choice of the right biomechanical parameters to improve is of great importance in the first step for a creation of a digital twin.

2.1 Analysis of the Workflow and the Specific Environment

Typical human activities in e.g. the automotive industry, aircraft production, logistics, retail, are e.g. handling loads, performing tasks at head height or above or assembling very small products. These and other tasks lead to different strains (e.g. with regard to the body region,). The analysis of such tasks is important, as it defines the starting point for interactions between human and technology in order to improve the quality of work and to relieve employees. Four main activities in particular need to be distinguished in industrial context: Lifting and carrying, working at and above head height, pushing and pulling or drilling and screwing. Depending on the activity, different parts of the body are stressed to different degrees [12]. On the basis of the identified task, various distinguishing characteristics can be derived for activities in industrial production and must be taken into account by introducing a exoskeletal system in the workflow of a company:

  • Dynamic and static activities: Distinction with regard to the speed of movement. On the one side, highly dynamic activities and on the other side, activities in static positions are distinguished [12].

  • Variance of tasks within the activity: For example, rotating workplaces with standing and sitting parts, and tasks at and above head height as opposed to monotonous work. The characteristics are the extent or ratio of secondary activities within a main activity [5].

  • Activities with and without components/tools: Depending on the activities, different objects or tools have to be used, whose handling and thus the regional stress can differ significantly (e.g. screwing with a screwdriver or screwing with a cordless screwdriver).

  • Weights to be handled: Differentiation with regard to the weights to be handled of parts, components, assemblies, workpieces or tools. The weights can vary from a few grams up to 30 kg.

  • Process forces/interaction forces: Amount of forces acting on the body during an activity. This does not include tool/component weights, but forces that act due to the work process, such as torques or contact pressure when assembling a component. during the assembly of a component.

  • Range of movement: Activities are not only carried out with different postures, but also in different ranges of movement. This is particularly important for occupational safety (e.g. collision with work equipment, falls) but also for the assessment of stresses, since small ranges of movement (e.g. assembly locations that are difficult to access) cause more forced postures.

2.2 Biomechanical Analysis

In order to determine the requirements and the exact need for the support of an exoskeletal system, the movement sequence of the workflow must be analyzed biomechanically beforehand. For this purpose, it is crucial which parameters can be examined. Biomechanical analysis usually includes different aspects of the interaction of the human with its environment e.g. the movement (kinematics), external forces and moments (kinetics) acting on the body or caused by its interaction with the environment, internal forces and also the muscle activity that cause voluntary body movement. The following parameters summarizes a selection of the most important values used for the evaluation of biomechanical effects for a physical support system – exoskeleton [11].

  • Body movement: Specific motion patterns such as joint angles, trajectories, dynamics (velocities, accelerations) with and without support.

  • Muscle activity: Muscle activity in percent of maximum muscle activity (%MVC, maximum voluntary contraction).

  • Cardiovascular/cardiopulmonary activity/metabolic effort: Heart rate, O2/CO2 relation, relative VO2 in ml/min/kg.

  • Force/torque (inertial, external): Joint reactions forces/torques, ground reaction forces, balance, center of pressure.

  • Individual perception: Comfort level, user acceptance, psychological aspects.

The most used biomechanical analysis in the evaluation of physical support systems (exoskeletons) of all listed methods in this section is the electromyographical analysis (EMG) [14]. This reference value is created by maximum voluntary contraction (MVC) measurements directly prior to the actual measurements. The general idea behind MVC measurements is that these will present themselves as a 100% contraction value. Any measurements will be presented as a percentage value in reference to the MVC [14].

2.3 Digital Human Model

Basically, the implementation of a digital twin requires suitable tools and a digital environment that can be configured. In order to be able to carry out dynamic biomechanical analyses for the prediction of relevant biomechanical parameters (e.g. internal forces, torque, muscle activities) acting on the human body Digital Human Model (DHM) are required. DHM software is a computer-aided design tool [15]. It can be evaluated from an ergonomics perspective using virtual simulation before making the real physical prototype. A few popular DHM software, which is commercially available include JACK, Sammie, Ramsis, Open Sim and the Biomechanics of Bodies (BoB) [10]. Some important previous research work [15] was conducted with a biomechanical modelling system, namely, the AnyBody Modeling System (AMS). This system provides the possibility to investigate the interaction of biomechanics at the musculoskeletal level. In such musculoskeletal models, structures like bones, tendons or muscles are modelled very detailed. AMS offers the possibility to simulate these models in interaction with its environment and to perform an inverse kinematic analysis.

To perform a simulation, motions and reactions normally have to be recorded from a subject (human) and transferred to the DHM. For recording movements, usually motion capturing are used. The entire human movement is captured with the help of a limited number of markers on the body via optical, three-dimensional kinematic camera [14]. Afterwards the position and orientation of the markers is transferred to the DHM with inverse kinematics to perform the simulation.

For this paper AMS was chosen because of the possibility to model additional mechanical variables (exoskeletal effect) in the simulation environment.

2.4 Process for the Prototypical Implementation of a Digital Twin

  • Step 1 – Analysis of the biomechanical need for physical support: According Sect. 2.1 the analysis of the relevant task(s) in the workflow must be carried out and pre-analyzed in detail to exclude the possibility that important factors are not taken into account.

  • Step 2 – Modeling of activities: After deriving the work activity to be investigated, the motion sequence must be recorded in detail using motion capture.

  • Step 3 – Consideration of existing exoskeletal systems: As a starting point for the developers, exoskeletons that are already available on the market should be analyzed for the application case. Already realized designs can be verified regarding their biomechanical effect. This effect can be incorporated in the simulation as a starting point for a possible physical support system that must be optimized.

  • Step 4 – Creation of the DHM: After selecting the appropriate DHM as described in Sect. 2.3, the digital twin must be parameterized according to size and weight of the users. The movement must be imported from the motion capturing and validated for further simulation.

  • Step 5 – Applying load and support force in the digital human model: The modelling of the mechanical parameters of the exoskeleton (in the form of external forces or moments with defined points of application on the human body) must be implemented in the simulation. For this virtual effect there is no need of a CAD model of the exoskeleton.

  • Step 6 – Setting different support characteristics: Input variables (forces, moments) defined in step 5 must be varied (parameter study) accordingly in order to define an optimal configuration of the physical support for the user. Other external variables (environmental parameters) must also be implemented in this step.

  • Step 7 – Evaluating the relevant biomechanical parameters – parameter study: Biomechanical parameters (output variables) must be evaluated (inverse dynamic analysis). Possible anomalies and correlations should be investigated - the optimal physical support for the specific user requirements must be identified.

3 Application Example: Overhead Lifting Task

To illustrate the application of a digital twin for an exoskeletal system evaluation, an example with corresponding parameter study has been provided in this paper. The aim of this evaluation is to determine the different effects of weight and support changes on muscle activity for the defined movement sequence. A common overhead work activity from the industrial sector was selected as a use case: Overhead (right arm) lifting activity. To investigate the effect of exoskeletal support when lifting a variable load, a model was built in the AMS to replicate the characteristics of the exoskeleton Lucy [9] as a starting point. Lucy is a shoulder exoskeleton and supports the abduction/elevation of the upper arms by means of a pneumatic actuator depending on the angle between the upper arm and the upper body (upper arm elevation angle) [9]. The model created in AMS generates a torque in the right shoulder depending on the angle between the humerus and thorax, which acts upwards. It therefore supports the lifting of the arm forwards (anteversion) and sideways (abduction). The following figure shows the movement sequence called the humerus-thorax elevation. This movement was recorded according step 2 in Sect. 2.4 with motion capture (Fig. 1).

Fig. 1
figure 1

Movement sequence of the overhead work activity (displayed in AMS)

Full size image

The subject stands upright at the beginning of the movement. The left arm hangs freely downwards. The muscles of the left arm are not examined in this study, which is why the movement of the left arm during the drilling activity is not described further here. The right arm is already pointing forward at the start of the movement. During the motion capturing recording, the test person held the drilling tool in his right hand. This is not shown in the simulation. The mass of the tool is changed in the different scenarios and acts at the center of the right palm. The test person moves the right arm evenly upwards until the tip of the tool reaches the point of action. In a realistic drilling process, the subject would now increase the pressure to perform the drilling. This detail is not simulated in this simulation for reasons of clarity and simplicity. Since different masses are defined as tool weight in the different scenarios, the same effects should be shown this way. This use case is limited to a consideration of muscle activities. Muscle activity is defined as the active state of the muscle in fractions of the maximum voluntary contraction. This means that at a muscle activity of 100%, the muscle has reached its theoretical load limit. A value greater than 100% is not possible in practice, but can occur in the simulation [13]. The muscle activities are measured as a representative parameter for measuring the relieving effect of the exoskeleton on the human musculoskeletal system. The maximum of the average muscle activity gives information about the unevenness of the effort. The further away the maximum is from the average mean value, the more irregular the load.

The parameter study is based on several scenarios, in each of them one parameter is changed and a comparison is generated. In the first comparison, two simulations are carried out in which the movement is simulated without exoskeletal support. The load in the first simulation is 0 kg and in the second simulation it is 2 kg. The simulation without load corresponds to the simple lifting of the right arm. These two scenarios are simulated, among other reasons, in order to obtain a reference for the results and to carry out a kind of plausibility check of the simulation. The assumption that muscle activity will increase with increasing load could be plausibly proven. In order to analyze the influence of the load, simulations with 4, 6 and 10 kg will be carried out sequentially. In the second comparison, the movement is simulated with the support of the assisting torque. The implemented torque curves (virtual torque applied to the shoulder hinge) are based on the exoskeleton Lucy and are shown in Fig. 2. The theoretical maximum assistance power for the exoskeleton Lucy is 12 Nm at a humerus-thorax angle of 90 degrees [9].

Fig. 2
figure 2

Support torque with maximum peak at 90 degree

Full size image

3.1 Results

In this use-case, only the most straining muscles in the shoulder were taken into consideration. For this purpose Deltoideus (anterior), Supraspinatus and Infraspinatus were analyzed in the situations. For a better understanding the structure of the shoulder muscles is illustrated in Fig. 3 (right side). It can be easily seen in Fig. 3 (left side) that in the part of the cycle where the applied torque is high, the effect on muscle activity is also high. It can also be seen that the effect on the Deltoid muscle is greater than on the Infraspinatus muscle. Especially the last third of the cycle, the exoskeletal support does not have such a large effect on the muscle activity of these two muscles.

Fig. 3
figure 3

Left: Muscle activities under varying support (torque) with 2 kg load, Right: Anatomy of the shoulder

Full size image

The results of all the different scenarios were summarized in Figs. 4 and 5. For this purpose, they are displayed in boxplots. The minimum, the lower quartile, the median, the upper quartile and the maximum are shown. For each muscle, a boxplot was created with one box per simulation scenario.

Fig. 4
figure 4

Muscle activity of the Supraspinatus (left) and Infraspinatus (right) under varying support and load

Full size image
Fig. 5
figure 5

Muscle activity of the Deltoideus Anterior under varying support and load

Full size image

The main task of the Supraspinatus muscle in this use case is abduction and external rotation of the upper arm, especially below an abduction angle of 15°. In Fig. 4 (left) the second box (0 Nm support, 2 kg load) and the fourth box (12 Nm, 2 kg load) are compared, it can be seen that the torque support has a clearly positive effect on muscle activity. The last three boxes (right in the diagram) show that an increase in load also increases the scatter of the data.

The Infraspinatus muscle is mainly responsible for the external rotation of the upper arm. The exoskeletal support has a different influence here than on the Supraspinatus muscle. It is noticeable in Fig. 4 (right) that in the second to fourth box (from the left side of the diagram) the upper quartile and the maxima are very close together. Here it is mainly the interquartile range that changes. This means that there is definitely a relief of the Infraspinatus through the exoskeletal support.

The Deltoideus Anterior muscle is largely involved in lifting the arm forward and is therefore the most important muscle in this activity. The greatest effect on muscle activity is therefore expected. This can also be easily seen in the 2nd to 4th box in Fig. 5 (from the left side). With successive increases in torque, all values decrease constantly. Increasing the load results in clearly higher muscle activity, whereby the scatter of the data increases also.

3.2 Discussion of the Results

In general, the simulation results show that a higher supporting torque leads to lower muscle activities in the Deltoideus Anterior, Infraspinatus and Supraspinatus. The lowest muscle activities were seen at lower loads and at greater supports. An important observation is that for this movement in this simulation, the increase in average muscle activity is approximately linear with a constant increase in load. However, it is not possible to predict whether this increase remains similar with further increases in load. The simulation is particularly suitable for looking at the muscles individually, different effects can be observed for the individual muscles. For example, in the Supraspinatus, the dispersion of the data increases when the load is increased. With increased torque, on the other hand, the dispersion remains similar, and the mean value of the activities decreases. The interquartile range (IQA) also remains similar, suggesting that the load remains consistently strenuous. For the Infraspinatus, the dispersion of the data increases with higher support torque but the mean value decreases. For the Deltoideus Anterior, the dispersion remains relatively constant with an increase in torque. The aim for an optimal system is to keep the dispersion as low as possible to keep muscle activity at a constant low level [13]. In general, the results confirm that muscle activity increases with higher load and decreases with higher support torque. For the downward movement of the arm, the support torque has less effect on muscle activity, as muscle-work has to be done against the system. It should be noted that the weight of the exoskeleton was not considered with the assumption that this would not have a major impact on the muscle activities considered here. Contact forces were also not taken into account for the use case presented. It is assumed that the contact forces have a greater influence on the subjective feeling of comfort when wearing a exoskeleton than on the measurable muscle activities. Another point of discussion is the way in which the support-torque is transmitted to the body or how exactly it acts. In the virtual model (DHM), it is a torque in the shoulder hinge without contact points.

4 Conclusion and Outlook

This paper has examined the possibility as well as the potentials of using a digital twin with regard to the evaluation of different characteristics of a physical support system. In the first approach, promising evaluation possibilities could be shown without using a real system. The digital twin provides a fast and agile way to investigate user-specific configurations and derive an optimal support setting, which is essential for the construction of a real exoskeleton. The next step will be to deduce the correct mechanical design of the exoskeleton based on the optimal support characteristics. For this purpose, mechanical (active/passive) elements have to be dimensioned to generate the necessary support. A possible validation of the design would be the import of the CAD of the exoskeletal model in the DHM.