Model Assessment of the Complex Workload of Harvester Operator

Abstract

This article deals with the complex evaluation of a working environment. The aim of this paper is to find a mathematical model for a comprehensive risk assessment of a harvester operator. The developed model is based on the mutual evaluation of dependent and independent variables. The selected independent (explanatory) input variables of the model are the equivalent sound pressure level, peak sound pressure level, operative temperature, and mental stress. The selected dependent variable is the riskiness of the work, which we determined by means of heart rate variability evaluation. Based on the underlying measured data, we compiled a mathematical model that seems to be perspective. This model takes into account quantitative information on risk factors that can be determined by objectivization, as well as qualitative information on the health of the assessed person.

1. Introduction

Operating a forest harvester is a complex task, requiring considerable time to master. The tasks performed by the harvester operator are well known; an operator is expected to: position the machine for harvesting, select and reach the tree to be felled, fell and position the stem to the processing (i.e., delimbing, bucking, scaling, and grading) location, and bunch the bucked logs, sorted by assortment, at the side of the trail so they can be collected by the forwarder [1]. Aside from productive operations, a harvester operator is often required to perform additional tasks, such as evaluating terrain conditions, identifying possible obstacles, or performing maintenance and small repairs on the machine.

The working environment in a harvester is demanding. Harvester operators are exposed simultaneously or sequentially to multiple occupational hazards. They affect the operator’s workload and well-being, thus influencing their performance and productivity. The majority of published studies are focused on a separate analysis and evaluation of various risk factors of the harvester operator’s working environment, such as noise [2,3], whole-body vibration [3,4], microclimate conditions [5], illuminance [6], physiological workload [7,8,9], and mental workload [10,11,12].

The main limitation of the aforementioned approach is that it does not consider the effects of simultaneous exposure to multiple risk factors. To remediate this, several authors have tried to find different methodological approaches based on multi-factor evaluations of the harvester operators’ work environment. Gerasimov and Sokolov [13] developed a method for the ergonomic evaluation of timber harvesting systems that is based on multi-criterial decision making using the Hodges–Lehman rule. In contrast, Jankovský et al. [14] based their assessment of selected risk factors (whole-body vibration, noise, microclimatic conditions, mental workload) on a combination of measured risk factor values and the perception of these factors by the harvester operators. Marzano et al. [15] proposed a methodology for the determination of an Ergonomic Conformity Index for forest machines, considering the aspects related to exposure to noise, vibration, thermal environment, and air quality. Hnilica et al. [16] outlined the possibilities for using the analytical hierarchy process for the complex assessment of the harvester operator’s work environment. Jankovský et al. [17] developed two generalized linear models that described the influence of the cumulative effects of work-related factors (noise, whole-body vibration, microclimatic conditions, and illuminance, among others) on increasing the harvester operator’s heart rate. Iftime et al. [18] reported results from an analysis on the synergistic effects of occupational hazards on forestry vehicle operators, as well as the incidence of professional and occupational disease caused by exposure to these risks. Yet another approach was developed by Oliveira et al. [19], who developed a method to evaluate the ergonomic performance of harvesters by using the Integrated Ergonomic Indicator, which is capable of correlating ergonomic variables simultaneously.

Performing a holistic evaluation for working environment with multiple exposures is a major challenge, given that most of the methods and tools available are based on exposure limits defined for single risk factors. Although research studies have been conducted by several authors, methodological approaches based on multi-factor evaluations are still insufficiently explored. As far as we know, no previous research has investigated the procedure that utilizes mathematical modelling techniques to assess the effects of simultaneous forest harvester operator exposure to multiple risk factors in a manner similar to the approach used in this study. The aim of this paper is to develop a mathematical model for the complex workload evaluation of a harvester operator’s working environment.

2. Materials and Methods

A simulation of the complex workload was conducted in a laboratory setting. A harvester simulator produced by Valmet was set up in an ergonomics laboratory, as can be seen in Figure 1. The room was lit by a combination of natural and artificial lighting (ten roof-installed fluorescent lights). The noise exposure in the lab varied, depending on the simulation activities on the harvester simulator, with the noise being transmitted to the participants through the simulator’s speakers. The laboratory space, with natural ventilation, was heated by conventional heating equipment (radiators).

The harvester simulator was controlled by the participants during data collection. During the simulated work, they were exposed to the working environment and the work itself. The group of participants consisted of 30 students (26 men, 4 women) aged 20–25 years. The participants were full-time students who operated the harvester simulator as part of their practical training. Each student operated the simulator for approximately 30 min. Occasionally, the measurements had to be excluded from further study due to scrambled signals from the devices, leading to a total sample size of 849 measurements.

On the harvester simulator, Valmet 911.4, we simulated the working process of a harvester operator (Figure 1) working in thinning operations of a spruce forest stand on level surface (or mild slopes up to 10% incline). The mentioned work process of the harvester was offered by the program. These were thinning operations with the possibility of changing the parameters of the terrain. Prior to the measurements, the trainees were instructed on how to work with the simulator and tested the operation of the machine. The simulation of harvester operation started at the same time as the measurements of the factors of the work environment. We selected the factors so that they would represent the actual work risk for harvester operators, based on their relevance and significance in mechanized forest harvesting. The following factors of the work environment were selected for measurements:

-

noise→personal noise dose meter Quest Technologies, model EDGE eg3 (Figure 2a),

-

microclimate→ambient air quality meter Testo 480 (Figure 2b),

-

mental loading→subjective assessment through the Meister questionnaire,

-

heart rate→Biofeedback 2000 x-pert (Figure 2c).

Figure 2. Measurement devices: (a) Quest Technologies, model EDGE eg3; (b) Testo 480; (c) Biofeedback 2000 x-pert.

Figure 2. Measurement devices: (a) Quest Technologies, model EDGE eg3; (b) Testo 480; (c) Biofeedback 2000 x-pert.

Noise was measured based on the standard STN EN ISO 9612:2010 [20]. We interpreted the results according to the current legislation, namely government regulation No. 115/2006 Coll. [21] and regulation No. 555/2006 Coll. [22]. The measurements of mental and sensorial loading of the trainees were conducted according to regulations of the Ministry of Health of Slovak Republic No. 67, which describes the approach of government bodies regarding limiting mental loading at work [23]. The microclimate condition measurements were performed according to Decree of the Ministry of Health of Slovak Republic No. 99/2016 Coll. [24], which describes the details of health protection against heat and cold loading at work. The parameters of the work environment included in the model as explanatory variables are shown in

Table 1. Independent and response variables.

Variable	Observed Parameters
Response	y	Risk coefficient
Independent	x1	Equivalent noise pressure level LAeq,T (dB)
	x2	Peak noise pressure level LCPk (dB)
	x3	Operative temperature to (°C)
	x4	Mental loading ML (-)

.

An operator’s physiological response (biofeedback) to workload was expressed as the change in their heart rates while operating the harvester simulator. The heart rate of the participants was measured simultaneously with the measurements of factors of the work environment by the MULTI module of the Biofeedback 2000 x-pert device (photopletysmograph), which was positioned on the temporal region of the participants‘ head. The device recorded the heart rate at a rate of 100 Hz. The heart rate measurements were divided into two steps for each participant: workload measurements and rest phase measurements. The workload measurements lasted for approximately 30 min. After the heart rate measurements started, participants started operating the harvester simulator; then, after 30 min, the heart rate recording was stopped and the participants stopped working with the simulator. After the workload measurements, the rest phase (control) measurements were conducted. The rest phase measurements lasted for approximately five minutes. After the work with the simulator ended, the participants were seated in an ergonomic armchair and rested, with minimal muscle movements. In some cases, they were silent. If that was uncomfortable for the participants, they held a calm conversation with the researcher. The purpose of these measurements was to establish a baseline heart rate for that particular day. A detailed description of the data gathering and processing method was described by [14,25].

We then constructed a binary response variable y, based on the heart rate of the trainees under workload. According to [26,27,28,29,30,31,32,33,34,35,36], heart rate correlates with all independent variables x₁ to x₄. Heart rate correlates to work-related loading and can be used to predict work-related health risks. Apart from the fact that heart rate correlates with many factors of the work environment, another advantage is that Slovak legislation has clearly defined decision-making criteria that can be implemented for assessing the occupational risk levels based on heart rate. The binary response variable entered the state “risk” or “1”, if the heart rate under loading or the difference of heart rate under loading from resting heart rate exceeded the limits stated in [37], based on criteria for assessing heart rate at work (

Table 2. Criteria for assessing the change in heart rate at work [37].

Age Group	Values of Shift Heart Rate per Minute
	Absolute Values		Increase Heart Rate above Baseline
	A Average Values	B Limit Values	C Average Values	D Limit Values
18 to 29	108	117	30	33
30 to 39	106	115	29	32
40 to 49	101	110	26	28
50 to 59	97	105	23	25
60 to 65	93	100	20	22

A—value determined to assess findings in a group of persons, if the baseline heart rate frequency value is not determined; B—value which may still be acceptable for the investigated person in the long term if the C value is not exceeded; C—maximum allowable value of heart rate frequency increase above the baseline that is long term acceptable for healthy individuals; D—maximum allowable value of heart rate frequency increase above the baseline, which must not be exceeded.

). Otherwise, the loading was considered “safe” and the binary response variable state was “0”. We measured heart rate by a Biofeedback 2000 x-pert (Figure 2c) device. Measurements of the physiological response (heart rate) to workload were performed according to the methodology described [25].

To assess the work environment from the view of complex loading, we analyzed the response interval of variable y from independent interval variables x₁ to x₄. Due to the range of the necessary input data, we verified the proposed method for the complex work environment assessment on trainees who operated a harvester simulator in a laboratory.

We designed a model capable of assessing the complexity of the factors of the work environment that would be as accurate as possible based on the data we gathered. The model was based on a forward stepwise regression and correlation analysis of our statistical sample. The resulting mathematical model was based on the analysis of the responses of the interval variable Y from independent interval variables X. Specifically, it was based on the correlations between the values of the risk factors and the value of the physiological response of the human body to work-related loading.

During simulations, we worked with a linear model of the following type (Equation (1)):

$$y=\left({\alpha }_1{x}_1\right)+\left({\alpha }_2{x}_2\right)+\left({\alpha }_3{x}_3\right)+\left({\alpha }_4{x}_4\right)+{\alpha }_5$$

(1)

where y is the coefficient of risk (legislation-set binary variable based on heart rate under workload),

• α_j are the coefficients of the significance of risk factors,
• x₁ is the equivalent noise pressure level L_Aeq,T (dB),
• x₂ is the peak noise pressure level L_CPk (dB),
• x₃ is the operative temperature t_o (°C),
• And x₄ is mental loading ML (-).

The outcome of this linear model shall be a function through which we should be able to predict the outcome at preselected risk factor Xj values.

3. Results

The mean normalized level of noise exposure of the trainees was L_AEX,8h = 75.84 dB. The outcome of noise analyses proved that working on a harvester simulator did not present an occupational hazard; only in one case was the lower action value of noise exposure according to Slovak legislation (80 dBA) [22] exceeded. Besides the normalized noise exposure level, we observed the peak noise level. The trainees were, on average, exposed to L_CPk = 103.19 dB, which is well under the Slovak legislation limits [22]. The total mean operative temperature in the laboratory was 24.27 °C. The data show no substantial temperature fluctuations, with temperature variability being approximately 2 °C, which was likely caused by the passive ventilation of the laboratory.

The mental loading survey was conducted on trainees who had no practical experience with the operation the machines, which should be considered when interpreting the results. The trainees considered working with the harvest simulator to be moderately overloading. The median score was 3 out of a possible 5 points. They also found the work moderately monotonous with a median score of 3 out of 5 points. The participants perceived non-specific loading to be the most challenging for them, with the highest absolute median score of 4 out of 5 points being reached in non-specific loading. This was partially caused by the highest number of questions dealing with this factor. Generally, we can state that working with a harvester simulator presented severe loading on the participants. Meister’s questionnaire is critically median, and it was exceeded in the factors of overloading, monotony, and non-specific loading.

The mean workload heart rate of the participants was 94.76 bpm. The mean minute rest phase heart rate of the trainees was 80.65 bpm. An important indicator was the difference between the workload and the rest phase heart rate, which was 18.9 bpm.

Based on the observations of the binary response variable, the total share of work considered “risky” during training on the harvester simulator was 23.67% on a statistical sample of 849 cases. We therefore considered more than 3/4 of the cases of the trainees’ work on be harvester simulator “safe” from the viewpoint of heart rate under workload or its change under loading. From the observation of health risk stemming from heart rate increase, we can say that:

-

from the view of mean absolute heart rate, the share of risky work was 23.09%,

-

from the view of marginal absolute heart rate, the share of risky work was 19.32%,

-

from the view of mean heart rate difference, the share of risky work was 19.32%,

-

from the view of marginal heart rate difference, the share of risky work was 18.61%.

Mathematical modeling (

Table 3. Linear regression model.

Model	R	R Square	Adjusted R Square	Std. Error of the Estimate	Change Statistics
					R Square Change	F Change	df1	df2	Sig. F Change
1	0.222 a	0.049	0.044	0.41593	0.049	8.760	5	843	0.000

^a Predictors: (Constant), ML (x₄), t₀ (x₃), L_CPk,T (x₂), L²_Aeq,T (x²₁), L_Aeq,T (x₁).

and

Table 4. Analysis of variance of the constructed mathematical model.

Model		Sum of Squares	df	Mean Square	F	Sig.
1	Regression	7.578	5	1.516	8.760	0.000 b
	Residual	145.836	843	0.173
	Total	153.413	848

^b Predictors: (Constant), ML (x₄), t₀ (x₃), L_CPk,T (x₂), L²_Aeq,T (x²₁), L_Aeq,T (x₁).

) showed that the coefficient of correlation between the binary response variable and all factors of the work environment was R = 0.222 and the coefficient of determination was R² = 0.049. Moreover, from

Table 5. Multiple regression and correlation analysis.

Model		Unstandardized Coefficients		Standardized Coefficients	t	Sig.	95.0% Confidence Interval for B
		B	Std. Error	Beta			Lower Bound	Upper Bound
1	(Constant)	13.640	2.540		5.370	0.000	8.654	18.626
	LAeq,T	−0.370	0.066	−5.149	−5.641	0.000	−0.499	−0.242
	L2Aeq,T	0.003	0.000	5.051	5.537	0.000	0.002	0.003
	LCpk,T	−0.003	0.008	−0.013	−0.392	0.695	−0.020	0.013
	to	0.017	0.020	0.029	0.844	0.399	−0.022	0.056
	ML	0.086	0.031	0.096	2.812	0.005	0.026	0.146

we can see the significance of particular factors. Factors L_Aeq,T and $L_{\mathrm{Aeq},T}^2$ and mental loading (ML) were significant.

In the next step, we verified the significance of the constructed mathematical model. Based on the outcomes of the test in Table 4, we found the model was significant, with (F = 8.76 a p = 0.000).

Starting from Table 3 and Table 5, we are aware of the results of the multiple regression and correlation analysis. The specific formula for the created model is (Equation (2)):

$$y=13.640-\left(0.370·L_{\mathrm{Aeq},T}\right)+\left(0.003·L_{\mathrm{Aeq},T}^2\right)-\left(0.003·L_{\mathrm{CPk},T}\right)+\left(0.017·t_0\right)+\left(0.086·ML\right)$$

(2)

The co-variances of the resulting coefficients of the mathematical model are shown in

Table 6. Co-variances of the model variables.

Model			ML	to	LCPk,T	L2Aeq,T	LAeq
1	Co-variances	ML	0.0009354	−0.0000171	−0.0000165	0.0000005	−0.0000786
		to	−0.0000171	0.0003972	0.0000012	0.0000004	−0.0000714
		LCPk,T	−0.0000165	0.0000012	0.0000690	−0.0000001	0.0000117
		L2Aeq,T	0.0000005	0.0000004	−0.0000001	0.0000002	−0.0000298
		LAeq	−0.0000786	−0.0000714	0.0000117	−0.0000298	0.0043119

, which shows that the co-variances between the individual independent variables were small and not statistically significant. Therefore, the mathematical model met the presumption of linear independence of the independent variables.

The linear regression model that describes the relationship between the factors of the work environment and the binary response variable can be further improved by including only the significant variables. The effect of individual significant factors of the work environment on the response variable can be seen in

Table 7. Stepwise multiple regression and correlation analysis.

Model	R	RSquare	Adjusted R Square	Std. Error of the Estimate	Change Statistics
					R Square Change	F Change	df1	df2	Sig. F Change
1	0.081 a	0.007	0.005	0.42420	0.007	5.573	1	847	0.018
2	0.199 b	0.039	0.037	0.41735	0.033	29.003	1	846	0.000
3	0.220 c	0.048	0.045	0.41565	0.009	7.943	1	845	0.005

^a Predictors: (Constant), L_Aeq,T; ^b Predictors: (Constant), L²_Aeq,T, L_Aeq,T; ^c Predictors: (Constant), L²_Aeq,T, L_Aeq,T, ML.

.

The stepwise analysis showed that adding significant variables into the model increased the coefficient of correlation and determination. For additional assessment, we selected model no. 3, on which we tested its significance. The outcomes of this test (

Table 8. Analysis of variance of the selected model.

Model		Sum of Squares	df	Mean Square	F	Sig.
3	Regression	7.427	3	2.476	14.330	0.000 d
	Residual	145.986	845	0.173
	Total	153.413	848

^d Predictors: (Constant), L²_Aeq,T, L_Aeq,T, ML.

) show that the selected mathematical model was suitable for complex work environment assessment (F = 143.3 and P = 0.000).

For the final version of the model, which is described in

Table 9. Multiple regression and correlation analysis of the selected model.

Model		Unstandardized Coefficients		Standardized Coefficients	t	Sig.	95.0% Confidence Interval for B
		B	Std. Error	Beta			Lower Bound	Upper Bound
3	(Constant)	13.564	2.348		5.776	0.000	8.955	18.173
	LAeq,T	−0.367	0.066	−5.099	−5.600	0.000	−0.495	−0.238
	L2Aeq,T	0.002	0.000	5.008	5.501	0.000	0.002	0.003
	ML	0.086	0.030	0.096	2.818	0.005	0.026	0.146

, we devised the following formula (Equation (3)):

$$y=13.564-\left(0.367·L_{\mathrm{Aeq},T}\right)+\left(0.002·L_{\mathrm{Aeq},T}^2\right)+\left(0.086·ML\right)$$

(3)

Formula (3) logically describes the presumed effects of individual factors of the work environment on the “riskiness” of work—increasing noise mental loading of the participants increased the binary response variable and therefore the riskiness of work. The resulting correlation and co-variances of the variables of the selected mathematical model are shown in

Table 10. Correlations and covariances of the variables in the selected model.

Model			LAeq,T	L2Aeq,T	ML
3	Correlations	LAeq,T	1.000	−0.999	−0.039
		L2Aeq,T	−0.999	1.000	0.034
		ML	−0.039	0.034	1.000
	Covariances	LAeq,T	0.0042913	−0.0000297	−0.0000788
		L2Aeq,T	−0.0000297	0.0000002	0.0000005
		ML	−0.0000788	0.0000005	0.0009295

.

4. Discussion

To create a suitable mathematical model, we must observe the statistical significance of the relationships of the model’s qualitative and quantitative parameters. The quantitative parameters observed through objective risk factors can be realized relatively easily, due to the availability of a large spectrum of standardized methods. However, the qualitative parameters should be based on information obtained through diagnosing the clinical symptoms (e.g., during a physical examination), direct observation of physiological variables (heart rate changes, bodily temperature, blood pressure, reaction time, etc.), or a questionnaire survey.

Regarding the form for the complex assessment of work environment itself, it is necessary to conduct a partial measurement of the individual risk factors that we want to observe. The importance of these kinds of assessments is documented by the large body of research focused on this problem. During all direct measurements, we focused only on unselected risk factors of the work environment, with which we verified the suitability of our approach and designed a new mathematical model for complex work environment assessment.

In the presented article, we proposed and assessed our proposed method of complex work environment assessment. This area is still relatively sparsely researched, compared to partial work environment assessments. In the literature, we did not come across this approach to work. The papers we found on the complex work environment assessment were largely focused on comparing two risk factors, such as mental loading and noise [38,39,40]. The experimental study of the effects of noise on the concentration of trainees by [38] showed that the subjective susceptibility to noise was the main factor responsible for the significant differences in concentration at task resolving. Hathaway [39] observed the effects of noise levels on concentration during work. A person exposed to noise on a long-term basis exhibits deterioration of mental and physical productivity, is unstable, irritable, unable to focus, their reaction time deteriorates, and they prematurely fatigue [40].

Similar to noise, operative temperature also affects physical and mental performance. Olsen [41] states that incorrectly set operated temperatures in the workplace leads to a decrease in physical and mental performance. Workers exert more energy to sustain homeostasis and therefore muscles and other tissues are not capable of optimal physical performance. The mental performance of an individual also decreases at high temperatures. The problem of complex work environment assessment was researched in studies such as [7,42,43,44,45,46,47,48,49,50,51,52,53,54].

Kapustová [45] assessed the work environment in a blacksmith’s workshop. She applied the coefficient method of complex work environment assessment. Authors in studies [46,48,55,56] applied statistical methods when they determined the efficacy of the multi-criterial assessment of the work environment. Piňosová et al. [55] elaborated a multi-criterial mathematical modeling and first-stage complex work environment assessment. Sablik [43] focused on a scoring method of a complex work environment assessment.

The proposed mathematical model in Equation (3) could serve for simulation, where, for example, with one fixed factor we could observe the development of other factors and their effect on occupational health and safety. The mathematical model for a complex work environment assessment was gradually filled with data gathered from measurements of individual risk factors in laboratory conditions.

After evaluating the selected risk factors (independent variables), which were measured and simulated, we estimated the relationship between the physiological response of the human body and workload. From laboratory measurements, we found that the intensity of the physiological response varied from one person to another. It is important to realize that the human body was affected by multiple factors, which could bias the assessment outcomes. These factors include, for example, family background, sleeping patterns, and other external factors. In our research, we did not consider these factors. It is therefore apparent that true complex assessment of the work environment is complicated.

From the gathered results and the analyses, we found that the proposed mathematical model could form the basis for further improvement. This model could serve for simulations, where, for example, for one fixed factor it is possible to monitor the development of other factors and their impact on health. However, it is necessary to continue studying this approach to complex work environment assessment and to verify its real-life harvesting. The constructed mathematical model can explain the variability of work-related risks, which means it can serve as a tool in complex work environment assessment. To improve the developed mathematical model, it is necessary to monitor the job from the perspective of risk factors and their impact on human health in the long term. Here, we see the possibility for cooperation with experts in the field of occupational health services and preventive occupational medicine. It will be important to monitor the long-term effects of risk factors (noise, mental stress, physical stress, microclimate, etc.) on the health and safety of workers.

Based on these statistical analyses we considered only those variables that had a significant effect on the response variable. These included the equivalent level of noise, its square, and mental loading.

During training on the harvester simulator, it is possible to simulate various risky situations during the training and better prepare the operator for operation. Of course, training in operating conditions in the field alone will not replace it, but it can simulate (show) risk factors that affect the well-being of the operator. Subsequently, during the simulation, it is possible to identify people who are unsuitable for this job position.

5. Conclusions

Ensuring a high standard of occupational health and safety is one of the most important elements of a successful and calm atmosphere in a workplace that stimulates its workers. The body of factors of the work environment acting upon the workers during their work, their intensity, and potential synergisms create a unique work environment and form its quality. As the production systems develop, so must the methods that assess their effects on the workers. One of the possible methods of such an assessment is the complex assessment of the work environment. In this paper, we designed and evaluated a multi-criterial model for the complex evaluation of the work environment. The proposed interactive mathematical model enables the simulation of real-life working conditions, where every partial factor of the work environment can be manipulated individually, with other factors remaining unchanged. The constructed model creates a basis for evaluating the effects of each factor on the overall quality of the work environment or studying the interactions between particular risk factors.