AI-Powered Computer Vision for Ergonomic Risk Assessment and Musculoskeletal Symptom Prevalence in Industrial Metal Polishing Operators

Abstract

Manufacturing polishing tasks involve repetitive movements and sustained postures that increase exposure to work-related musculoskeletal disorders (WRMSDs). This study presents an intersectoral validation of the ergonomic assessment methodology applied to industrial metal polishing operators that considered sociodemographic, anthropometric, and health variables. This study surveyed 41 workers using the Nordic Musculoskeletal Questionnaire and assessed a subsample of 27 workers using the REBA method through AI-based computer vision. Symptom prevalence was highest in the neck (82.9%), shoulders (70.8%), lower back (68.3%), and wrists/hands (65.9%). Using a computer-vision AI-based tool to analyse posture, the REBA method identified moderate (70.3%), high (26.0%) and very high (3.7%) WRMSD risks for the upper arms, neck, and trunk, respectively, with women showing greater susceptibility. Spearman correlation analysis revealed significant associations between age, gender, health perception, and musculoskeletal risks. The findings confirm the ergonomic assessment method’s applicability and reliability for ergonomic risk assessment in industrial polishing tasks, emphasising the need for targeted interventions adapted to gender and age profiles to mitigate occupational hazards. The results support a non-intrusive assessment approach suitable for industrial deployment and for prioritising targeted, worker-stratified ergonomic interventions.

1. Introduction

Industries are facing a growing sociodemographic challenge characterised by an ageing workforce and declining birth rates, a phenomenon known as the “demographic winter”. At the European level, the low fertility rate combined with an increasingly pronounced ageing population today could jeopardise Europe’s economic and social sustainability in the coming decades [1].

This situation demands a review of occupational health policies and human resource management, aiming to adapt working conditions to the capabilities of an ageing workforce and to promote retention and training strategies [1,2].

However, occupational health in manufacturing industries has been significantly affected by the rise in symptoms and work-related musculoskeletal disorders (WRMSDs), especially in sectors involving repetitive manual tasks and high physical demands [3,4].

Heavy-duty industrial sectors have shown a high prevalence of musculoskeletal symptoms, particularly in the spinal region (lumbar, thoracic and cervical), which are associated not only with biomechanical demands, such as inappropriate posture and repetitive tasks, but also with organisational and psychosocial factors, including long working hours and work-related stress [5]. Another study of porcelain-manual work revealed high rates of musculoskeletal symptoms in the spine (neck and lower back) and elbows, confirming that musculoskeletal symptoms are a common concern in industrial environments and that sociodemographic and occupational factors help to support their occurrence [6].

Furthermore, studies performed in Portugal also reveal a high prevalence in the manufacturing sector. Over 50% of workers in the needle industry report complaints in the lower back (54%), the neck (42%), shoulders (39%), ankles/feet (38%) and wrists/hands (36%) over the past 12 months, with a higher incidence among women and older workers [7]. In the transport-metalworking sector, workers also reported musculoskeletal symptoms, mainly in the spine and upper limbs, with a risk of developing work-related musculoskeletal disorders [8]. This high prevalence is associated with risk factors common to all these sectors, namely repetitive movements, inappropriate postures, manual handling and work-related stress, which are key factors in manufacturing industries, leading to musculoskeletal injuries in the upper limbs [7,8]. Tuners, and metalworking operators, on the other hand, report more complaints due to static postures and vibrating work tools, respectively [7,8].

Recent studies, such as those by Alves et al. [3,4], highlight the importance of detailed ergonomic assessments and worker-centred approaches for the prevention and mitigation of these risks, particularly within the context of the transition to Industry 5.0, where worker well-being and inclusion are key priorities [9].

In parallel with these demographic shifts, the landscape of ergonomic assessment is undergoing a digital transformation. Recent literature highlights a paradigm shift from traditional observation-based methods towards AI-driven, markerless motion capture systems. These computer vision technologies offer a non-intrusive alternative to wearable sensors, enabling continuous and objective risk monitoring without interfering with worker productivity [10,11]. Furthermore, the integration of these digital tools within the Industry 5.0 framework is increasingly recognised as a critical enabler for sustainable manufacturing, allowing for data-driven interventions that are tailored to individual worker profiles rather than generic population averages [12].

Recent studies have demonstrated that computer vision and machine learning can automate ergonomic risk assessment by estimating body joint angles and corresponding RULA scores in real time, thereby overcoming the reliance on and limitations of manual observation [13]. Furthermore, the application of wearable sensor systems with machine learning has enabled kinematic monitoring and the prediction of injury risk in dynamic activities [14]. Recently, AI-driven Digital Twins incorporating computer vision have enabled the integration of pose estimation, object detection and the surrounding environment for comprehensive optimisation of the workspace and for immersive, faster, dynamic and real-time ergonomic interventions [15]. All these technological strategies open doors for occupational health.

Therefore, integrating ergonomic aspects with demographic challenges constitutes an essential approach for ensuring productive, safe, and inclusive environments in modern industry. This work focuses on replicating the methodology applied by Alves et al. [3,4], utilising standardised questionnaires such as the Nordic Musculoskeletal Questionnaire (NMQ) and advanced posture recognition technologies aided by artificial intelligence for dynamic assessment of musculoskeletal risks. To this end, a group of workers performing identical polishing tasks within the same industrial sector was studied and analysed to evaluate and validate the methodology, making it viable and reproducible in other industrial contexts.

Furthermore, intersectoral comparison was assessed between samples stratified by gender and age clusters, allowing the identification of specific vulnerabilities and the targeting of customised ergonomic interventions. Consequently, this study advances the hypothesis that the ergonomic assessment methodology, combining NMQ and AI-based posture analysis, produces consistent risk profiles across different cohorts within the same industrial sector, while remaining sensitive enough to detect specific vulnerabilities related to gender and age.

2. Materials and Methods

To replicate and validate the methodology previously applied by Alves et al. [3,4] in the polishing sector of the precision metalworking industry, this investigation followed and replicated the same methodological approach. The earlier studies characterised the sociodemographic and symptomatologic profiles of workers through the Nordic Musculoskeletal Questionnaire (NMQ), adapted and validated for the Portuguese population [16], alongside a sensorless ergonomic assessment tool based on computer vision and artificial intelligence algorithms, TuMeke, to evaluate the risk of work-related musculoskeletal disorders (WRMSDs) [17].

TuMeke is a sensorless and computer vision-based application that automates the Rapid Entire Body Assessment (REBA) methodology. The system employs advanced pose estimation algorithms, comparable to frameworks such as MediaPipe or OpenPose, to track skeletal key points from standard 2D video input. By reconstructing the subject’s 3D posture geometry, the software automatically calculates joint angles relative to gravity and adjacent body segments, determining the precise flexion, extension, and rotation values required for the REBA scoring algorithm. This automated approach eliminates the need for intrusive wearable sensors or manual protractor measurements, thereby reducing intra- and inter-observer variability while maintaining the standardised scoring criteria of the original REBA method [4,17,18].

This investigation, therefore, takes the form of an intersectoral and methodological validation, since it was conducted in the same operational polishing sector, which is highly dependent on manual processes and considered high risk for WRMSDs, within the same type of metallurgical industry, but with a distinct sample of operators. Replication in an identical context makes it possible to assess the robustness, consistency, and reproducibility of the methodology defined by Alves et al. [3,4].

The conduct of an intersectoral validation within a context closely aligned to the original study is essential to ensure the robustness, validity, and reliability of the measurements obtained, particularly when applied to different samples performing identical tasks within the same industrial sector. Reliability, defined as the instrument’s ability to consistently reproduce the relative ranking of participants across repeated assessments, constitutes an essential requirement for providing consistent and comparable data [19]. However, reliability alone does not guarantee absolute agreement between repeated measures, which is critical for distinguishing genuine changes from measurement error or tool sensitivity [19]. Methodological validity demands that the instrument not only be reliable but also accurately and meaningfully measure the intended construct, thereby justifying rigorous adaptation and validation of the questionnaires and ergonomic assessment systems employed [20]. This study’s replication of the methodology in a distinct group of operators undertaking the same tasks in the same industrial environment facilitates not only the evaluation of result reproducibility but also strengthens external validity by confirming the applicability and consistency of the findings beyond the initial sample.

The study was carried out in July 2025 at an industrial unit in the metalworking sector, located in central Portugal, specialising in polishing and in the manufacture of luxury metal products and jewellery. The company employs between 50 and 200 operators and is recognised not only for its high level of technical expertise but also for its social and corporate responsibility towards its clients and other stakeholders. As in the studies [3,4], the operational department selected was the polishing division, characterised by repetitive manual tasks requiring consecutive movements of the upper limbs and trunk, usually performed in a seated position. All participants in the study held no supervisory roles and were exclusively engaged in the manual polishing of small metal pieces, an activity involving the application of varying levels of force throughout the working day.

Data collection was conducted through two complementary approaches. Firstly, the sociodemographic and musculoskeletal symptomatology questionnaire, “Ergonomics, Wellbeing, and Health in the Industrial Context” (Supplementary Materials S1), adapted from previous literature and structured in two sections, was administered [3]. This questionnaire incorporated personal and professional data alongside the Portuguese version of the NMQ [8,16,21]. The NMQ recorded the presence of fatigue, discomfort or pain (FDP) in nine anatomical regions, analysed across three reference periods: the previous 12 months, the previous 7 days, and functional limitations in the past 12 months, i.e., prevented from doing work (PDW). Secondly, a sensorless ergonomic assessment was performed using the TuMeke application, which estimates 3D posture through smartphone video recordings of the polishing task [17,18]. This new technological approach has been a key focus since the start of the project, as it was essential to ensure that companies’ productivity was not affected. For this reason, the study was designed to have the least possible impact on normal operational functioning. Workers were recorded at three different times of the day (morning, midday break/lunch, and afternoon), and the footage was analysed with the software to generate ergonomic risk indicators based on established methodologies, namely the REBA [22]. To minimise observer bias, all participants were informed beforehand and given a short acclimation period before recording.

Following the methodology of Alves et al. [3,4], the sample was grouped by age categories: young (18–34 years), intermediate (35–44 years), and ageing (≥45 years). However, in this study, the ageing group was subdivided into two clusters: ageing workers (45–54 years) and senior workers (55–64 years). Workers were randomly selected within each category available in the sector. Thus, although this study follows the same general methodological framework as our previous work, it was conducted independently; that is, with a new sample of operators, collected over a different company and period, and with an improved age group design (i.e., greater stratification within the group of elderly workers). So, this stratification enabled comparative and stratified analyses of symptom prevalence and ergonomic risks by age and gender, ensuring replication of the original methodological design.

Finally, the collected variables were harmonised for further statistical treatment. Normality test Shapiro-Wilk indicated non-normal distribution and small samples (Supplementary Materials S2), and therefore non-parametric correlation tests (Spearman’s coefficient) were applied to assess associations between sociodemographic information, musculoskeletal symptomatology, and ergonomic risk indicators.

The study was approved by the Ethics Committee of the University of Beira Interior (reference code CE-UBI-Pj-2024-001). All participants provided informed consent, including authorisation for image and video recording. Confidentiality of personal data was ensured through encryption and restricted access granted solely to the research team.

3. Results

The initial sample consisted of 41 workers, none of whom declined voluntary participation, resulting in a 100% response rate. It is important to note that, in this case study and during the analysis phase, none of the operators were excluded for having a body mass index (BMI) below 18.5 kg/m² or above 40 kg/m². Therefore, it was not necessary to apply this exclusion criterion, thereby eliminating the potential significant impact of BMI on work-related musculoskeletal disorders.

3.1. Sociodemographic, Professional and Health Characterisation of the Total Sample

This section presents the sociodemographic and professional characteristics of the workforce, including age, gender, marital status, cohabitation with children, educational level, years of service, work schedules, working hours, and equipment used. Furthermore, it outlines health-related habits, lifestyle factors, and anthropometric measures, as well as the prevalence of self-reported symptoms associated with physical demands and musculoskeletal health, thereby providing an overview of the workers’ general health status and work-related musculoskeletal complaints.

3.1.1. Sociodemographic and Professional Characteristics

The entire sample comprised 41 polishing operators with a mean age of approximately 43 years (σ = 9.157), ranging from a minimum of 20 years to a maximum of 59 years. In terms of gender, the sample consisted of 28 females (68.3%) and 13 males (31.7%), with a mean age of approximately 45 years for females (σ = 8.342) and 36 years for males (σ = 7.933).

Regarding sociodemographic characteristics, namely marital status, most workers were married (53.7%), with significant gender differences: females were predominantly married (71.4%), whereas males were mostly single (46.2%) or in a common-law partnership (30.8%). Concerning cohabitation with children under the age of 16, the total sample was evenly distributed, with 51.2% of individuals not living with children in this age group, and 46.4% living with children under 16. This trend was also balanced across genders, with 50% of females not cohabiting with children under 16 years, while the proportion was slightly higher among males, where 53.8% did not live with children under 16. As for educational qualifications, most workers, that is, 58.5%, had completed secondary education (10th to 12th grade). This was followed by those who had completed basic education (5th to 9th grade) with 19.5%, and technical–vocational education with 12.2%.

Regarding the data reflecting the professional situation of the operators, the greatest percentage of operators have been with the company for less than 5 years (48.8%), followed by those with a length of service between 5 and 10 years (31.7%). The average seniority in the company is 6.1 years (σ = 4.287). In this context, the average seniority for males is 8.4 years (σ = 5.893), a higher value than for females, whose average is 5.1 years (σ = 2.968). The average weekly working hours were 40 h for all operators. Similarly, work schedules were consistent and the same for all operators, as this company/industry operates only regular working hours. Regarding the equipment used, 56.1% of workers used polishing equipment, followed by 31.7% of operators who used lapidary machines.

Table 1. Sociodemographic and Professional Data (n: frequency; %: percentage; NR: No Response).

	Total Sample		Female Gender		Male Gender
	n	%	n	%	n	%
Gender
Female	28	68.3
Male	13	31.7
Age Clusters
18–34	7	17.1	2	7.1	5	38.5
35–44	19	46.3	12	42.9	7	53.8
45–54	7	17.1	6	21.4	1	7.7
55–64	8	19.5	8	28.6	-	-
NR	-	-	-	-	-	-
Marital Status
Single	8	19.5	2	7.1	6	46.2
Married	22	53.7	20	71.4	2	15.4
Divorced	5	12.2	4	14.3	1	7.7
Common Law	6	14.6	2	7.1	4	30.8
NR	-	-	-	-	-	-
Children < 16 Years Old
No	21	51.2	14	50	7	53.8
Yes	19	46.3	13	46.4	6	46.2
NR	1	2.4	1	3.6	-	-
Educational Qualifications
Primary Education	1	2.4	1	3.6	-	-
Basic Education	8	19.5	5	17.9	3	23.1
Secondary Education	24	58.5	16	57.1	8	61.5
Technical–Professional Education	5	12.2	3	10.7	2	15.4
Higher Education	2	4.9	2	7.1	-	-
Postgraduate Higher Education	-	-	-	-	-	-
NR	1	2.4	1	3.6	-	-
Seniority in the Company
<5 years	20	48.8	15	53.6	5	38.5
5 to 10 years	13	31.7	10	35.7	3	23.1
>10 years	6	14.6	2	7.1	4	30.8
NR	2	4.9	1	3.6	1	7.7
Working Hours/Week
<40	-	-	-	-	-	-
40	41	100	28	100	13	100
>40	-	-	-	-	-	-
NR	-	-	-	-	-	-
Work Schedules
Regular	41	100	28	100	13	100
Shifts	-	-	-	-	-	-
Work Equipment
Vitax/Polishing	23	56.1	13	46.4	10	76.9
Lapidary Machine	13	31.7	11	39.3	2	15.4
Other	5	12.2	4	14.3	1	7.7

presents a summary of the values corresponding to the sociodemographic and professional variables, considering the entire sample and its distribution by gender.

3.1.2. Lifestyle, Health Habits and Complaints of Work-Related Musculoskeletal Disorders

About the polishing operators, 82.9% identified their right hand as dominant and 14.6% declared using both hands with equal dominance. This trend is even more pronounced among females, with 89.3% reporting right-hand dominance, compared to males, where only 69.2% reported right-hand dominance. Among male operators, the proportion declaring ambidexterity was higher, at 30.8%. Considering weekly physical exercise, 65.9% of workers reported not engaging in physical activity, with a higher prevalence among females (71.4%) than males (53.8%). However, this difference between the genders is less marked in the male gender, despite the overall low physical activity levels. Regarding sleep habits, 53.7% of operators reported sleeping between 7 and 8 h, while 42.9% slept between 4 and 6 h, with similar values across genders. Concerning self-reported overall health status, most workers rated their health as reasonable (51.2%), followed by good (19.5%), and with equal percentages providing ratings at the extremes of very good and poor (9.8%). Using the Nordic Musculoskeletal Questionnaire (NMQ), weight and height data were collected for the calculation and classification of body mass index (BMI). For all operators, the mean weight was 71.7 kg (σ = 14.054), with a minimum of 50 kg and a maximum of 107 kg. Considering gender, the mean weight was 66.5 kg (σ = 9.571) for females and 82.8 kg (σ = 15.968) for males. The mean height was 1.67 m (σ = 0.090), being 1.62 m (σ = 0.067) for females and 1.76 m (σ = 0.062) for males. Consequently, the mean BMI for the total sample was classified as overweight, with a mean BMI of 25.6 (σ = 3.955). Specifically, for females, the mean BMI was 25.2 (σ = 3.946), and for males it was 26.5 (σ = 3.969), both indicating overweight.

Table 2. Health and lifestyle habits (n: frequency; %: percentage).

	Total Sample		Female Gender		Male Gender
	N	%	n	%	n	%
Dominant Hand
Right	34	82.9	25	89.3	9	69.2
Left	1	2.4	1	3.6	-	-
Both	6	14.6	2	7.1	4	30.8
BMI Situation
Normal Weight	24	58.5	19	67.9	5	38.5
Overweight	11	26.8	5	17.9	6	46.2
Obesity Class I	6	14.6	4	14.3	2	15.4
Obesity Class II	-	-	-	-	-	-
Physical Exercise
No	27	65.9	20	71.4	7	53.8
Yes	14	34.1	8	28.6	6	46.2
Sleep Habits
Between 1 and 3 h	1	2.4	1	3.6	-	-
Between 4 and 6 h	18	43.9	13	46.4	5	38.5
Between 7 and 8 h	22	53.7	14	50	8	61.5
More than 8 h	-	-	-	-	-	-
Health Perception
Excellent	2	4.9	-	-	2	15.4
Very Good	4	9.8	2	7.1	2	15.4
Good	8	19.5	4	14.3	4	30.8
Reasonable	21	51.2	19	67.9	2	15.4
Deficit	4	9.8	3	10.7	1	7.7
NR	2	4.9	-	-	2	15.4

presents a summary of the data corresponding to the health and lifestyle habits information, considering the entire sample and its distribution by gender.

The existence and prevalence of FDP, and/or work-related musculoskeletal disorders were assessed using the Portuguese version of the Nordic Musculoskeletal Questionnaire (NMQ). According to this questionnaire, the analysis was conducted for different anatomical regions, specifically the spine (individually examining the cervical, thoracic, and lumbar sections), the upper limbs (including shoulders, wrists/hands, and elbows), and the lower limbs (analysing hips/thighs, legs/knees, and ankles/feet). Furthermore, the data were analysed for two different periods, first the previous 12 months and then the previous 7 days, as well as regarding any limitation in performing usual work tasks (Figure 1).

The graphical representation in Figure 1 shows that the percentage of complaints relating to FDP in the last 12 months was highest in the spinal regions and the upper limbs, especially in the shoulders and wrists/hands. The neck was identified by operators as the area with the highest prevalence of complaints (82.9%), followed by the shoulders (70.8%), lower back (68.3%), wrists/hands (65.9%), and the upper back (63.4%).

Concerning complaints reported by operators in the past 7 days, unlike the reports of the previous 12 months, the upper limbs were the anatomical area with the highest prevalence, with 46.4% for the shoulders and 39.1% for the wrists/hands. Next were the vertebral regions, with 39% of complaints in the lower back, and 34.1% in both the upper back and the neck. Regarding limitations in performing work tasks during the previous 12 months, the highest incidence was found in the spine, especially in the neck (12.2%), followed by the shoulders and lower back (9.8%). It is noteworthy that all anatomical areas studied, except for the elbows and ankles/feet, registered limitations in work performance. Thus, comparing the percentages of FDP complaints in different anatomical regions and the corresponding work limitations suggests a relationship, namely, the inability to perform certain tasks may be associated with the high prevalence of musculoskeletal complaints in these regions, related to musculoskeletal disorders or injuries.

The intensity of FDP experienced by the operators in these anatomical areas was then analysed, with the severity rated as mild, moderate, severe, or unbearable (Figure 2). The intensity classification shown in Figure 2 indicates that the most frequently reported by operators was moderate for all anatomical regions, except for the wrists/hands, where the most frequent classification was intense (24.4%). However, the highest values for moderate intensity were observed in the spinal regions, specifically 39% in the neck, 31.7% in the lower back, and 26.8% in the upper back; and in the upper limbs, particularly the shoulders (31.7%). It is also important to highlight the moderate intensity reported for the knees (29.3%) and hips/thighs (19.5%). Notably, in these anatomical regions, the second most prevalent classification was intense intensity, indicating a higher degree of seriousness for the FDP reported by these workers. In the wrists/hands specifically, there was a very similar proportion of moderate (22%) and intense (24.4%) complaints.

3.1.3. Intra-Sectoral Comparison of Industries

According to the sociodemographic data, the polishing section of the company examined in this study exhibits a higher prevalence of female workers compared to males, whereas the polishing section of the company studied by Ref. [3] displays a more balanced distribution, with 45% female and 55% male. The predominant age group in this study is between 35 and 44 years old (46.3%), followed by 18 to 34 years (17.1%) and 45 to 54 years (17.1%), while the company assessed by Alves et al. [3] has a higher representation in the 18 to 34 age group (39.3%) and a more homogeneous distribution among other age categories. Regarding marital status, this study features a lower proportion of single workers (19.5%), with most being married (53.7%), whereas Ref. [3] includes 44.3% single and only 36.4% married, highlighting a notable difference in the social profiles of the groups. The predominant educational attainment in the polishing department of this industry is upper secondary education (58.5%), followed by basic and technical-professional education. In Ref. [3], upper secondary is also predominant (43.6%), but there is greater diversity in basic and higher education levels.

Considering professional and work habits data, in the industry analysed in this study, 48.8% of operators have less than 5 years of seniority with the company, comparable to Alves et al. [3], where 71.4% have less than 5 years’ seniority, indicating a higher proportion of recent operators in that polishing department. In both groups, the majority work a 40-h week (100% in this study and 85.7% in Ref. [3], all working regular shifts (100% in this case and 85.7% in Ref. [3]. Regarding equipment, operators in this study mostly use polishing machines (56.1%), while a greater proportion in Alves et al. [3] use them (82.1%), with a smaller share operating lapidary machines and other tools.

Additionally, with respect to health habits and lifestyle, both groups exhibit predominance of right-hand dominance (82.9% in this group of workers and 62.9% in Ref. [3]. The majority present normal weight according to BMI (58.5% in this study and 52.1% in Ref. [3], though physical exercise is more prevalent among workers in Alves et al. [3] (57.9%) compared to those in this investigation (34.1%). Regarding sleep patterns, both worker groups show similar trends, with most sleeping between 7 and 8 h, comparable to those studied by Ref. [3]. Self-perceived general health among operators in this industrial case is predominantly rated as reasonable (51.2%), a more pessimistic assessment than that of Alves et al. [3], who reported higher percentages of good (47.1%) and very good (25.7%) perceptions.

Moving on to reported symptoms of FDP over the past 12 months, there was a predominance in the cervical, wrist/hand, and lumbar regions in both groups, although the values were slightly higher in this group of workers for almost all anatomical zones, especially for neck symptoms (82.9% vs. 73.6% in Ref. [3]) and shoulders (70.8% vs. 56.4% in Alves et al. (2024)), except for wrists/hands (65.9% vs. 72.8% in Ref. [3]) and the lower back (68.3% vs. 69.3% in Ref. [3]), which were slightly lower. In the acute period analysis (7 days), operators in this study maintained a higher prevalence in relation to Alves et al. [3], except in the neck and lower back. In terms of severity and intensity of symptoms, this study demonstrates a slight tendency towards a higher proportion of reports of moderate to intense pain across all anatomical areas compared to Alves et al. [3]

3.2. Comparison of the Total Sample Data with the Selected Study Sample for Ergonomic Analysis

The next phase of the study consisted of randomly selecting operators from each age cluster. To ensure an equal number of operators evaluated from each age range, the limiting factor was set by the maximum number of operators present in the smallest age groups within the overall sample. Thus, the limiting age clusters were 18 to 34 years and 45 to 54 years, each of which included only 7 operators. Consequently, the final sample comprised 27 workers, with 7 from each age group, except for the 45 to 54 years cluster, where only 6 operators were available to participate at this stage. The sample for the ergonomic assessment study represents approximately 66% of the total polishing operators. All recordings were carried out after analysing the full sample of operators involved in the polishing task, previously included and described in this study, having also provided informed consent and met the established inclusion criteria. The reduction in the number of participants from 41 to 27 was based on the same strategy of balancing participants by age group used in the previous study, ensuring that the groups were of the same size for a more balanced comparison between age categories and thus enabling methodological replication and validation.

3.2.1. Study Sample Data of Sociodemographic, Health and Anthropometry

The mean age of the sample was 43 years (σ = 10.673), comprising 19 female operators (70.4%) and 8 male operators (29.6%), with mean ages of 47 years (σ = 8.935) and 34 years (σ = 8.799), respectively. Across age groups, there was a higher proportion of male workers in the youngest cluster and a higher proportion of female workers in the older clusters. Regarding marital status, most operators were married (55.6%); however, when considered by gender, most male operators were single (37.5%) or in a common law partnership (37.5%), whereas most female operators were married (73.7%). As for cohabitation with children under 16 years, 55.3% of the sample did not live with children in this age group, a trend more pronounced among males, where 62.5% did not cohabit with children under 16 years, compared to 52.6% of females. Most operators have secondary education as their level of education, which is also the level of education most attained by both genders.

With respect to professional characteristics, 40.7% of workers had less than 5 years of seniority, closely followed by 31.7% who had been with the company between 5 and 10 years. Furthermore, the greatest prevalence of higher seniority, i.e., more than 10 years in the company, was found among men (25%), whereas the lower seniority categories had a higher prevalence among women.

Moving to lifestyle and health information, specifically hand dominance, 81.5% of workers reported right-hand dominance, with a greater prevalence among females (89.5%) and a lower prevalence among males (62.5%). However, men reported a higher rate of ambidextrousness (37.5%) compared to women (5.3%). For physical activity, 63% of operators did not engage in physical exercise, especially women (73.7%), in contrast to men, who reported higher rates of participation in physical activity (62.5%). Regarding sleep habits, most workers reported a sleep pattern of 7 to 8 h, observed in both genders. However, sleep patterns of 4 to 6 h were more common among females.

The mean BMI for the sample was 25.5 (σ = 4.030), classified as overweight, with 26.1 (σ = 4.156) for men and 25.2 (σ = 4.058) for women. Nevertheless, women presented a higher percentage of normal weight (68.4%), whereas men had a higher percentage of overweight (50%). By age groups, all showed mean values in the overweight category, since 25.4 (σ = 3.804) in the 18–34 cluster, 25.3 (σ = 4.736) in the 45–54 cluster, and to 27.0 (σ = 4.493) in the 55–64 cluster, except for the 35–44 age cluster, with a mean value at the threshold for normal weight, 24.1 (σ = 3.477). Cases of obesity were recorded, with a maximum BMI of 34.7 in the senior cluster (55–64), 33.1 in the youngest cluster (18–34), and 32.9 in the ageing cluster (45–54).

Perception of general health was most frequently rated as reasonable (51.9%) and good (22.2%). Among females, reasonable perception predominated (68.4%), whereas among males, good perception was more frequent (37.5%).

The patterns observed in this sample were consistent with and like the trends previously described above for the total sample (Figure A1).

As previously explained, the sample under study was randomly selected from the total sample, while ensuring a balanced age distribution across the different groups, specifically 18–34, 35–44, 45–54, and 55–64 years. In the initially analysed total sample, limiting age clusters were observed, with the 35 to 44 years group being the only cluster with a higher concentration of operators. Therefore, in the current study sample, operators from this age cluster were randomly selected to achieve a uniform distribution across the four age clusters.

Regarding gender distribution, some variations were recorded, but the percentage trends were preserved in all age clusters, maintaining a similar proportion and gender trend to that of the total sample. The analysed sociodemographic and health data show consistency and similarity between the study sample and the total sample, as well as between genders, thus preserving the same percentage trends. Although not significant, slight differences between genders were found when comparing the study sample and the total sample, specifically concerning divorce rates in marital status data, where men presented a higher percentage than women in the study sample, contrary to the total sample; similarly, in the BMI classification for class I obesity, women showed a higher percentage than men, contrary to the total sample.

Next, a comparison was made of the prevalence of work-related musculoskeletal disorder (WRMSD) symptoms perceived by the operators, as well as the intensity of self-reported symptoms, between the study sample and the total sample.

Regarding FDP in the last 12 months (Figure 3), operators in the study sample reported more symptoms in the following body areas: 81.5% in the neck, 74.1% in the lower back, and 74% in the shoulders and wrists/hands. These results showed similarities between the total sample and the study sample across all analysed anatomical regions. Nonetheless, a tendency towards slight percentage increases was observed for the shoulders, wrists/hands, lower back, and knees. The remaining body regions remained approximately constant or showed a tendency towards lower percentages. Gender analysis also demonstrated similar trends between the total sample and the study sample, with females reporting a higher prevalence of symptomatology than males.

Analysis by age groups similarly revealed consistent patterns and trends between the samples. In the study sample, the regions with the highest prevalence of complaints were the neck: 100% for the 55–64 age cluster, 85.7% for the 35–44 cluster, and 83.3% for the 45–54 cluster; wrists/hands: 100% for the 45–54 cluster, 85.5% for the 55–64 cluster; and the lumbar region: 100% for the 55–64 cluster, 71.4% for the 35–44 cluster. Overall, symptomatology data maintained the same trend across all age groups between samples, with higher percentages of self-reported symptoms observed in the older age clusters.

Similarly, an analysis was conducted of the occurrence of FDP symptoms in the last 7 days (Figure 4). Overall, the symptomatology reported in the last 7 days by operators in the study sample was like that of the total sample, with a tendency towards slightly higher percentages. When analysed by gender and age groups, the trends remained similar, exhibiting the same symptomatic patterns.

Additionally, the limitation in performing work tasks over the past 12 months was analysed and compared (Figure 5). The results showed equivalent patterns between the total sample and the study sample, except for the neck, shoulders, and wrists/hands regions, which presented higher percentages in the study sample. In contrast, the upper and lower back and knees showed lower percentages. It is important to note that, for the hips/thighs, no task limitations were noted in the study sample during the past 12 months. Gender-wise, the same trend was maintained across all body regions, except for the knees, where no limitations in task performance were recorded for males. Similarly, no limitations were observed in the study sample for the knees (contrary to the total sample in the 45–54 age cluster), with the rest of the age group analysis following the same patterns and trends.

Finally, a comparison was made of the intensity of pain reported by operators between the total sample and the study sample (Figure A2). This analysis showed similar results across all body parts, both by gender and by age clusters. Overall, the results reveal a higher prevalence of moderate and intense pain in the anatomical regions of the spine and the upper limbs.

3.2.2. Intra-Sectoral Comparison of Industries

The sociodemographic, anthropometric, and musculoskeletal symptom data of polishing operators were analysed in two distinct studies: one conducted in this investigation and another by Ref. [4]. The sample in this study showed a slightly higher mean age of 43 years compared to the Alves et al. [4] sample, which had a mean age of 40 years, with a female predominance in the present study (70.4%) and a male predominance in Ref. [4] (55%). The sample in this study was slightly older than that of Alves et al. [4]. Regarding industry/company seniority, both studies indicated a higher presence of workers with less than 5 years’ seniority, although Ref. [4] identified a higher proportion (66.7%) in this group compared to 40.7% in the current study.

Analyses of BMI showed similar means for both samples, 25.5 in this study and 25.6 in Ref. [4], indicating a tendency towards overweight. A greater prevalence of extreme obesity cases was observed in Alves et al.’s sample [4], particularly in younger age clusters, whereas this study also revealed extreme obesity cases in older age clusters (ageing and senior clusters).

Concerning the prevalence of symptoms related to WRMSDs, both studies reported greater incidence of FDP in the neck, wrists/hands, and lumbar regions, with symptomatology increasing proportionally with age and a predominance of complaints among females. This study indicated a tendency toward higher prevalence of symptoms in the shoulders and wrists, as well as greater functional limitations in these areas, compared to Ref. [4]. Moderate to severe pain intensity was similarly reported in both samples, highlighting the spine and upper limbs as the most affected regions.

In conclusion, the results demonstrate a consensus and resemblance between the two samples regarding the sociodemographic, anthropometric, and symptom profiles, pointing to consistency in the ergonomic and health conditions of operators within this industrial sector, particularly in the polishing department. The minor discrepancies observed, especially in gender composition and specific symptom incidence, may reflect differences in sampling and data collection periods, underscoring the importance of ongoing assessments to better understand occupational risk factors in these industrial environments.

3.3. Ergonomic Risk Assessment for WRMSD Prevention

To conduct the physical ergonomic analysis of the operators and the respective assessment of the risk of developing work-related musculoskeletal disorders (WRMSDs), video recording and the application of a sensorless system based on artificial intelligence algorithms integrated into a mobile application were used. This new methodology allows ergonomic analysis without the need for additional physical sensors, as their application was deemed unfeasible due to the impossibility of interrupting the production process, the high associated costs, and the time required for calibration and adjustment of the equipment [23,24].

3.3.1. Ergonomic Analysis and WRMSD Risk Assessment

Using the mobile application where recordings of the operators were uploaded, and which employs the REBA ergonomic assessment methodology, it was possible to evaluate the risk levels for developing work-related musculoskeletal disorders (WRMSDs) (Figure 6). For the study sample, a high prevalence of moderate risk for developing WRMSDs was revealed, i.e,70.3%, followed by 26% at high risk and 3.7% at very high risk.

Considering the analysis by gender, this assessment showed significant differences, with greater risk observed among women, although both genders presented a high percentage of moderate risk, specifically 63.2% in women and 87.5% in men. Women recorded a higher percentage of high risk (31.5%), whereas men presented lower percentages (12.5%). It is noteworthy that only the female gender exhibited 5.3% at very high risk, indicating a greater propensity for developing WRMSDs among female workers.

Regarding the age groups, differences were observed in the risk of developing WRMSDs related to the polishing task across the various age clusters. All age groups predominantly showed a high percentage of moderate risk: 85.7% in the young workers group (18–34 years old), 57.1% in the intermediate age group (35–44 years old), 50% in the ageing worker group (45–54 years old), and 85.7% in the senior worker group (55–64 years old). In addition, both the 18 to 34 and 55 to 64 age clusters presented 14.3% high risk of developing WRMSDs. More markedly, the 35 to 44 age cluster showed 28.6%, and the 45 to 54 cluster 50% at high risk. The most critical situation was centred on the 35 to 44 age cluster, with 14.3% at very high risk of developing WRMSDs.

Following the comprehensive analysis and assessment of the risk of developing WRMSDs in polishing operators, focusing on a whole-body evaluation, the REBA score results were used to conduct a segmented assessment by anatomical regions (Figure 7). This stratification was enabled by the detailed report provided by the TuMeke application, which, in addition to delivering the overall REBA score, presents the analysis broken down by body segments in accordance with the methodology’s criteria. This feature allows for the identification and interpretation of WRMSD risk levels in each body part, facilitating a more precise assessment aimed at targeted interventions.

The body regions of the neck, trunk, and legs showed higher percentages of low risk levels, followed by lower percentages of negligible risk. Conversely, the lower arm and wrists exhibited higher percentages of negligible risk, followed by low risk percentages. Notably, the body region with the highest risk of developing WRMSDs was the upper arm, presenting moderate risk levels. It is thus concluded that, overall, the body regions with the greatest WRMSD risk were the upper arms, followed by the neck and legs.

Comparing the symptoms reported by operators, previously identified as body areas with the most FDP, with the ergonomic risk assessment by body region, only the upper arms, neck, and trunk can be related as areas of risk for developing WRMSDs. Although the wrists were reported as one of the areas of greatest discomfort, they were not evaluated as a high-risk area in the ergonomic assessment.

The risk analysis and assessment across different body areas was also conducted, comparing genders. Generally, risk percentages followed the same trend in both genders when compared with the total study sample, except for the neck and legs, which showed only low risk (100%) for males. It is important to highlight that for all body regions, women exhibited slightly higher risk percentages, indicating a greater susceptibility to developing WRMSDs compared to men.

Similarly, the risk analysis was carried out considering different age groups (Figure 8). It can be stated that there is no specific pattern of risks associated with any age cluster, but the trends align with those of the overall study sample. However, the highest risks for developing WRMSDs are linked to the upper arm region (moderate risk), especially for the ageing and senior age clusters. Additionally, for all body regions except the neck, the ageing age cluster (45 to 54 years) presents the highest risks of developing WRMSDs.

3.3.2. Intra-Sectoral Comparison of Industries

Once again, a comparative analysis was conducted of the WRMSD risk levels derived from ergonomic assessments using the REBA methodology, comparing the data from this study with that of Ref. [4].

For the entire body, both studies indicated a high prevalence of moderate risk for developing WRMSDs, with Ref. [4] showing a higher moderate risk at 88.3%, compared to 70.3% in this study. However, regarding high and very high risk levels, this study revealed 26% and 3.7%, respectively, while Alves et al.’s study [4] reported only 15% high risk, with no very high risk recorded.

In the gender-stratified analysis, women presented higher risks of developing WRMSDs in both studies. Females in this study showed 63.2% moderate risk, 31.5% high risk, and 5.3% very high risk, whereas males showed 87.5% moderate risk and 12.5% high risk, with no very high risk recorded. In Ref. [4], moderate risk was 81.5% for women and 94% for men, with high risk at 18.5% for women and only 3% for men, highlighting greater female susceptibility.

When stratified by age clusters, both studies indicated predominance of moderate risk across all ages, reaching 100% among young workers (18–34 years) in Alves et al.’s [4] and 85.7% in the current study’s sample. The highest percentages of high risk were noted in the intermediate and ageing age groups, particularly in this study, where up to 50% high risk was observed in the 45 to 54 age group. Very high risk reached 14.3% in the 35 to 44 cluster in the total sample and 5% in the intermediate clusters (35–44 years) in Ref. [4], underscoring the vulnerability of more experienced workers.

In summary, both studies reaffirm the high ergonomic risk faced by polishing operators, especially among women and the intermediate and ageing age clusters, highlighting the need for ergonomic interventions that consider differentiated risk profiles by gender and age.

3.4. Correlation Analysis

Spearman’s rank correlation coefficient is a non-parametric measure that assesses the strength and direction of a monotonic relationship between two variables based on their ranks rather than raw values. Unlike Pearson’s correlation, which captures linear associations and is sensitive to outliers and normality assumptions, Spearman’s coefficient is more robust in the presence of outliers and is suitable for nonlinear but monotonic relationships. It ranges from −1, perfect negative monotonic association, to +1, perfect positive monotonic association, with 0 indicating no monotonic relationship [25,26].

3.4.1. Correlation Analysis of Study Sample

To begin with, a Spearman correlation was performed for the entire sample under study (Supplementary Materials S3), as outlined in Figure 9. For the total sample, the age variable showed a strong negative correlation (at the 0.01 level) with the gender variable, indicating that with increasing age, there is a tendency for operators to be female. Similarly, age also demonstrated a strong negative correlation (0.01 level) with the variable overall health perception, suggesting a trend towards more negative self-assessments of health as age increases. Additionally, the age variable exhibited positive correlations (0.05 level) with several variables related to musculoskeletal symptomatology reported by workers, particularly in the neck, lower arms, wrists, and trunk regions. This indicates an increase in reported FDP among older workers. Lastly, the age variable showed a negative correlation (0.05 level) with the REBA score for the legs, indicating that older operators tend to be exposed to lower musculoskeletal risks in the legs.

Regarding the gender variable, it showed a strong positive correlation (0.01 level) with the self-reported health perception variable, indicating that male workers tend to evaluate their health more positively. Conversely, gender also showed a strong negative correlation (0.01 level) with the variable for leg symptoms, meaning that male workers are less likely to report symptoms in the legs. The marital status variable showed a positive correlation (0.05 level) with BMI, suggesting that individuals with a higher BMI are more likely to be in a more advanced marital status, i.e., not single. Furthermore, educational qualifications were positively correlated (0.05 level) with lower arm symptomatology, indicating that workers with higher education levels report more symptoms in the lower arms.

Moving on to the analysis of the final sociodemographic variable, namely professional status, the variable seniority in the company was positively correlated (0.05 level) with BMI, suggesting that workers who have been with the company longer tend to have higher BMI levels. Similarly, seniority also showed a positive correlation (0.05 level) with the REBA score for the wrists, indicating a tendency for longer-seniority employees to exhibit higher risks of developing musculoskeletal disorders in the wrists. Additionally, seniority in the company presented a negative correlation (0.05 level) with self-perceived health, meaning that more aged operators tend to evaluate their health more negatively.

Next, an analysis was conducted of the health and lifestyle variables. The hand dominance variable showed a strong positive correlation (0.01 level) with BMI, indicating that operators who are ambidextrous tend to have higher BMI levels. The variable for physical exercise showed a strong positive correlation (0.01 level) with self-reported health perception, revealing that workers who engage in physical activity tend to evaluate their general health more positively. Moreover, physical exercise correlated negatively (0.05 level) with musculoskeletal symptom reports in the upper arms, wrists, and legs, and exhibited strong negative correlations (0.01 level) with REBA scores for the whole body, as well as for the upper arms, lower arms, and wrists. This indicates that workers who exercise regularly are less likely to report symptoms in these areas and have an even lower risk of developing work-related musculoskeletal disorders (WRMSDs) across these regions.

The variable for sleep habits showed a strong positive correlation (0.01 level) with self-perceived health, indicating that operators who sleep longer tend to evaluate their health more positively. This variable also showed negative correlations (0.05 level) with symptom reports in the lower arms, trunk, and wrists, suggesting that workers who sleep more tend to report fewer symptoms of FDP in these regions.

Self-reported health perception was also strongly negatively correlated (0.01 level) with musculoskeletal symptoms in the neck, wrists, and trunk, indicating that operators with better health perceptions tend to report fewer symptoms in these areas. Furthermore, health perception also showed a negative correlation (0.05 level) with REBA scores for the whole body and the lower arms, and a strong negative correlation (0.01 level) with REBA scores for the lower arms and wrists, suggesting that as operators rate their general health more positively, there is a tendency for lower WRMSD risks for the whole body, upper limbs, and particularly the lower arms and wrists.

Considering the musculoskeletal symptoms reported by operators, strong positive correlations (0.01 level) were found between neck symptoms and symptoms in the wrists and trunk, meaning that those who report neck symptoms also tend to report symptoms in the wrists and trunk. Likewise, symptoms in the upper arms were positively correlated (0.05 level) with symptoms in the wrists, indicating a tendency for individuals with upper arm symptoms to also report wrist discomfort. Additionally, wrist symptoms were strongly positively correlated (0.01 level) with trunk symptoms and also showed a positive correlation (0.05 level) with the REBA wrist score, indicating that workers who experience more wrist symptoms also tend to report trunk symptoms and are at greater risk of developing WRMSDs in the wrists. Likewise, trunk symptoms were strongly positively correlated (0.01 level) with leg symptoms, revealing that workers reporting discomfort in the trunk are also more likely to report symptoms in the legs.

Finally, the REBA score for the whole body was strongly positively correlated (0.01 level) with the REBA scores for the upper and lower arms, suggesting that operators with higher risks of WRMSDs in the arms are also at greater risk across the whole body. Furthermore, there was a strong positive correlation (0.01 level) between neck and trunk REBA scores, indicating a trend whereby operators at high musculoskeletal risk in the neck are also at elevated risk in the trunk. Similarly, a strong positive correlation (0.01 level) was found between REBA scores for the upper and lower arms. Conversely, the REBA score for the wrists showed a negative correlation (0.05 level) with the REBA score for the legs, suggesting that operators with a higher risk of WRMSDs in the wrists tend to have a lower risk in the legs.

Subsequently, the Spearman correlation analysis of the sample by gender was performed (Supplementary Materials S4), as illustrated in Figure 10. A strong positive correlation (0.01 level) was observed between the sociodemographic variables of age and seniority in the company, exclusively among males. In other words, older men tended to have longer seniority within the company. Also exclusive to the male group, age was negatively correlated (0.05 level) with workers’ health perception, indicating that older male operators tended to evaluate their health more negatively. Furthermore, among male participants, age demonstrated strong positive correlations (0.01 level) with musculoskeletal symptoms reported in the neck and wrists, suggesting that older men were more likely to report discomfort in these regions. For females, age was positively associated (0.05 level) with BMI, showing that older women tended to present higher BMI levels.

For women, marital status was negatively correlated (0.05 level) with reported upper arm symptoms, indicating that single women reported more frequent upper arm complaints. For men, educational qualifications were positively associated with REBA scores for the lower arms, suggesting that male workers with higher educational qualifications exhibited a greater risk of developing WRMSDs in the lower arms.

Regarding cohabitation with children under the age of 16, male operators who had longer seniority in the company were more likely to cohabit with younger children (positive correlation, 0.05 level). In addition, for men, this variable was negatively correlated (0.05 level) with health perception, revealing that male workers living with children tended to evaluate their health more negatively. Among men, cohabitation with children under 16 years also showed a positive correlation (0.05 level) with reported wrist symptoms and with REBA scores for the lower arms and wrists, indicating that these workers were more likely to report wrist discomfort and to face increased risks of developing WRMSDs in these anatomical regions.

With respect to seniority in the company, female operators only displayed a positive correlation (0.05 level) between seniority and BMI, showing that longer-seniority women tended to present higher BMI. For men, seniority in the company was strongly and negatively correlated (0.01 level) with health perception, meaning that older male workers with longer service were more likely to evaluate their health negatively. Additionally, male seniority was positively correlated (0.05 level) with reported symptoms of FDP in the neck, wrists, and trunk. Furthermore, older men also displayed a positive correlation (0.05 level) with REBA scores of the lower arms and wrists, indicating greater musculoskeletal risk in these areas.

Regarding hand dominance, correlations were identified solely among male workers. Hand dominance demonstrated a strong positive correlation (0.01 level) with BMI, indicating that ambidextrous male operators exhibited higher BMI levels. These workers also showed strong negative correlations (0.01 level) with exercise and sleep habits, suggesting that ambidextrous men tended to exercise more regularly and to experience longer sleep durations. However, a negative correlation (0.05 level) was observed between hand dominance and health perception, revealing that ambidextrous workers were more likely to evaluate their health negatively. Moreover, hand dominance was positively correlated (0.05 level) with REBA scores for the upper arms, lower arms, and wrists, signifying that ambidextrous men tended to face greater risks of WRMSDs in these anatomical regions.

Considering BMI, among males, there were strong negative correlations (0.01 level) between BMI and both exercise and sleep habits. This indicates that male workers with higher BMI levels were those who did not engage in physical exercise or had shorter sleep durations, suggesting that good health practices, including physical activity and healthy sleep habits, are linked to lower BMI. For men, BMI was also positively associated (0.05 level) with REBA scores for the upper arms, suggesting that a higher BMI is linked to increased risk of musculoskeletal disorders in this area. Conversely, among females, BMI was positively correlated (0.05 level) with musculoskeletal risk in the neck, indicating that women with higher BMI levels exhibited a greater risk of WRMSDs in that region.

Regarding physical exercise, for men, this variable was strongly and positively correlated (0.01 level) with sleep habits, indicating that male workers who engaged in physical activity tended to also have longer sleep durations. Physical exercise was further positively correlated (0.05 level) with health perception in men, meaning that male workers who exercised were more likely to evaluate their health positively. Among women, physical activity displayed a strong negative correlation (0.01 level) with upper arm symptoms, indicating that physically active female operators reported fewer upper arm complaints. Additionally, for women, physical activity was negatively associated (0.05 level) with global REBA scores, suggesting that female workers engaging in physical activities faced lower overall musculoskeletal risk. Among men, physical exercise was negatively correlated (0.05 level) with REBA scores for the upper and lower arms, showing reduced WRMSD risk in these regions. Both genders demonstrated a negative correlation (0.05 level) between physical exercise and REBA scores for the wrists, indicating that workers engaging in exercise have a lower risk of WRMSDs in the wrists.

Considering sleep habits, for men, this variable was positively associated (0.05 level) with health perception, indicating that those who slept longer tended to evaluate their health more positively. For women, sleep habits were negatively correlated (0.05 level) with lower arm symptoms, meaning that women with shorter sleep durations tended to report more symptoms in this region. For men, sleep habits also correlated negatively (0.05 level) with REBA scores for the upper arms, lower arms, and wrists, implying that shorter sleep durations were linked to greater musculoskeletal risks in these areas.

Regarding health perception, women who assessed their health more positively reported fewer neck symptoms (strong negative correlation, 0.01 level). Among men, health perception was negatively correlated (0.05 level) with wrist symptoms and REBA scores for the lower arms, indicating that men with better health perception reported fewer wrist problems and were exposed to lower risks of WRMSDs in the lower arms. For both genders, health perception was strongly and negatively correlated (0.01 level) with trunk symptoms and negatively correlated (0.05 level) with REBA scores for the wrists. Thus, workers who evaluated their health more positively tended to report fewer trunk symptoms and presented reduced risks of WRMSDs in the wrists.

For male operators, a positive correlation (0.05 level) was observed between reported neck symptoms and symptoms in the upper arms and wrists, indicating that those experiencing neck issues also reported complaints in these regions. Additionally, a positive correlation (0.05 level) was found between wrist symptoms and trunk complaints among men. For women, a strong positive correlation (0.01 level) was found between neck and trunk symptoms, showing that female workers with neck complaints reported trunk discomfort as well. For women, positive correlations (0.05 level) were identified between lower arm and upper arm symptoms, and between trunk and leg symptoms, suggesting clustered reporting of these complaints among female operators.

Finally, when evaluating musculoskeletal risks, the female group demonstrated stronger associations. Strong positive correlations (0.01 level) were identified between the overall REBA score and REBA scores for the upper arms, lower arms, and trunk, indicating that female workers at greater risk of WRMSDs in any of these regions also tended to present higher overall musculoskeletal risk. Among women, strong positive correlations (0.01 level) were also observed between REBA scores for the neck and lower arms, suggesting combined regional risks. Positive correlations (0.05 level) were identified between REBA scores for the wrists and trunk, and a negative correlation (0.05 level) was found between wrists and legs, revealing that women at greater musculoskeletal risk in the wrists also tended to have higher trunk risks but lower risks in the legs. By contrast, males displayed only a strong positive correlation (0.01 level) between REBA scores of the lower arms and wrists, meaning that male workers exposed to higher risk in the lower arms also exhibited higher musculoskeletal risks in the wrists.

As a final stage of analysis, the Spearman correlation was conducted by age groups (Supplementary Materials S5), as illustrated in Figure 11. Operators of intermediate age (35–44 years) showed a positive correlation (0.05 level) between gender and company seniority. This indicates that male operators within this intermediate age group tended to have longer seniority in the company. Younger workers (18–34 years) also revealed correlations with gender: a positive correlation (0.05 level) between gender and general health perception, meaning that younger male operators reported better self-assessments of their health; and a perfectly negative correlation (0.01 level) between gender and musculoskeletal risk in the lower arms, indicating that men in this age range were at minimal risk of developing WRMSDs in the lower arms.

Among the younger operators, there was a positive correlation (0.05 level) between marital status and REBA scores for the trunk, showing that younger workers who were no longer single tended to be at greater risk of WRMSDs in the trunk. By contrast, the ageing group (45–54 years) presented a strongly positive correlation (0.01 level) between marital status and health perception, indicating that workers in this age group who were married or in long-term unions evaluated their health more positively. Furthermore, these workers exhibited perfectly negative correlations (0.01 level) between marital status and reported symptoms in the upper arms and wrists, meaning that ageing operators who were not single were less likely to report symptoms in these regions. They also showed a negative correlation (0.05 level) between cohabitation with children under 16 and BMI, indicating that ageing operators without young children at home tended to exhibit lower BMI levels.

Considering company seniority, senior operators (55–64 years) presented a positive correlation (0.05 level) between tenure and BMI, suggesting that senior operators with longer service tended to have higher BMI levels. Operators of intermediate age displayed a negative correlation (0.05 level) between seniority and symptoms reported in the legs, showing that employees in this group with longer seniority tended to report fewer leg complaints, such as FDP. These operators also displayed a negative correlation (0.05 level) between seniority and REBA scores for the trunk, indicating that intermediate-age operators with longer seniority were less exposed to musculoskeletal risk in this region. Conversely, younger workers presented a strongly positive correlation (0.01 level) between seniority and trunk REBA scores, meaning that younger employees with comparatively longer service were more likely to face increased musculoskeletal risks in the trunk.

With respect to hand dominance, only the intermediate-age cluster displayed correlations, namely a positive correlation (0.05 level) between hand dominance and BMI. This suggests that operators in this group who used both hands for work tasks tended to exhibit higher BMI. Meanwhile, the senior cluster revealed a positive correlation (0.05 level) between BMI and REBA scores for the lower arms, meaning that older operators with higher BMI were more exposed to musculoskeletal risks in the lower arms.

Among younger workers, physical exercise correlated positively (0.05 level) with health perception, indicating that younger employees engaging in physical activity were more likely to evaluate their health positively. This group also showed a perfectly negative correlation (0.01 level) between physical exercise and symptoms in the lower arms, meaning that young operators who exercised did not report lower arm symptoms. Similarly, for ageing operators, a perfectly negative correlation (0.01 level) was detected between physical exercise and reported upper arm symptoms, showing that workers in this age group who practised physical activity did not report symptoms in the upper arms.

In relation to sleep habits, younger operators exhibited negative correlations (0.05 level) between sleep habits and REBA scores for the whole body, upper arms, and wrists. This suggests that younger workers with better sleep habits, i.e., longer sleep duration, faced lower overall musculoskeletal risk as well as reduced risks in the upper arms and wrists. With the same pattern, and more pronounced in the intermediate age group, sleep habits presented a perfectly negative correlation (0.01 level) with REBA scores for the wrists, showing that intermediate-age workers with longer sleep duration tended to have reduced musculoskeletal risks in the wrists.

For general health perception, ageing operators revealed perfectly negative correlations (0.01 level) with symptoms reported in the neck and trunk. This indicates that older workers who evaluated their health more positively were those who did not report symptoms in these regions. Younger operators also showed negative correlations (0.05 level) between health perception and symptomatology in the wrists, trunk, and legs, indicating that young workers who self-assessed their health more favourably reported fewer complaints in these regions. Furthermore, this group presented a negative correlation (0.05 level) between health perception and lower arm REBA scores, meaning that younger operators who assessed their health more positively were also at lower musculoskeletal risk in the lower arms. Intermediate-age workers showed a negative correlation (0.05 level) between health perception and REBA scores for the upper arms, suggesting that those with more positive health evaluations had lower musculoskeletal risks in the upper arms.

Regarding symptoms of FDP, the younger and intermediate groups (18–44 years) displayed a perfectly positive correlation (0.01 level) between symptoms in the neck and upper arms, suggesting symptom clustering in these body regions. A similar pattern was found in the ageing group, with a perfectly positive correlation (0.01 level) between reported neck and trunk symptoms.

Finally, for the extreme age groups, i.e., younger and senior workers, overall body musculoskeletal risks correlated positively (0.05 level) with REBA scores for the upper arms, indicating that workers with higher exposure in the upper arms also tended to have greater overall musculoskeletal risk. Moreover, among younger workers, overall musculoskeletal risk was positively correlated (0.05 level) with REBA scores for the trunk, suggesting that younger operators with musculoskeletal risks in the trunk tended to present overall body risks as well. Intermediate-age workers showed a strongly positive correlation (0.05 level) between REBA scores for the upper arms and lower arms, meaning that those with increased musculoskeletal risk in the upper arms also tended to present risks in the lower arms. Conversely, the senior cluster displayed a negative correlation (0.05 level) between REBA scores for the wrists and legs, showing that older operators at risk of WRMSDs in the wrists were less likely to face risks in the legs.

3.4.2. Intra-Sectoral Comparison of Industries

To better understand the complexity of associations between sociodemographic factors, lifestyle and health variables, musculoskeletal symptomatology, and ergonomic risks among polishing operators, a comparative Spearman correlation analysis was conducted between the present investigation and that of Alves et al. [4]. Both samples demonstrated strong correlations between health perception, musculoskeletal symptoms reported by operators, and ergonomic risk indices assessed using the REBA method. In general, both studies revealed that more negative self-assessments of health were associated with a higher prevalence of FDP among operators, accompanied by increased risk indices for developing WRMSDs.

Nevertheless, each sample displayed specific patterns. In the present study, the associations between age and reported symptoms were more pronounced, with older workers not only tending to report poorer health but also presenting a greater number of symptoms and higher exposure to risks across different body regions. Seniority within the company was correlated with higher BMI and musculoskeletal risks in specific areas, particularly in the wrists, suggesting cumulative effects of prolonged task exposure on ergonomic burden. In contrast, Ref. [4] reported that the strongest correlation concerned more experienced operators, i.e., those with longer company tenure, who tended to evaluate their health negatively and simultaneously exhibited heightened ergonomic risks across the whole body, with emphasis on the upper limbs. This finding reinforces the role of health perception as an indirect marker of occupational risk and highlights its potential utility for identifying more vulnerable subgroups within the workforce.

The gender-specific analyses in both studies underlined a consistent pattern of higher risk and symptom prevalence among women, with positive correlations observed between ergonomic exposure and reported musculoskeletal complaints. Conversely, male operators displayed a lower tendency to report symptoms and were associated with lower ergonomic risk indices, reflecting negative correlations between more positive health perception and the presence of musculoskeletal symptoms.

When stratified by age group, the findings suggested that younger workers tended to underestimate both symptoms and ergonomic risks. In some cases, this was evidenced by positive correlations between better health perception and higher ergonomic exposures, reflecting limited risk awareness. Among middle-aged and older operators, however, the correlations were more consistent: poorer health evaluations were associated with greater symptom reporting and higher ergonomic risks. This supports the hypothesis that occupational maturity is linked to a clearer perception of the severity of workplace risks.

In summary, the findings converge on the importance of multidimensional checking of workers, integrating sociodemographic characteristics, health perception, and detailed ergonomic assessments. The observed correlations emphasise the need for preventive and monitoring strategies, particularly targeting groups more vulnerable due to age, gender, or length of exposure. This approach may contribute to a deeper understanding of risk mechanisms in physically demanding and repetitive work environments.

4. Discussion

The results of this study, which focused specifically on metal polishing operators, revealed, using the NMQ, a higher prevalence of symptoms in the neck (82.9%), shoulders (70.8%) and lower back (68.3%) regions. Furthermore, using computer vision-AI technology, moderate-to-high REBA risks were identified, which are consistent with recent literature on repetitive industrial tasks, whilst also revealing some specific characteristics.

Using Spearman’s correlations, critical risk clusters for developing WRMSDs were identified among the polishing operators in the study, modulated by gender and age clusters. For example, age is associated with a higher incidence of symptoms in the spinal and wrist regions, with workers reporting poorer perceived health. It could also be said that the female gender is more vulnerable, particularly in terms of the risk of developing musculoskeletal disorders in the spine and upper limbs, whilst the risk for the male gender is exacerbated by seniority in the company and living with children. Furthermore, while young operators face a higher risk due to their seniority in the company, older workers face a higher risk due to a higher BMI and symptoms in multiple regions. It is important to highlight that physical exercise and adequate sleep emerge as cross-cutting protective factors, that is, for all workers, reducing musculoskeletal symptoms in the upper limbs and whole-body REBA risks. This reinforces the idea that gender- and age-specific priority interventions are necessary, namely, mandatory training, task rotation and ergonomic adjustments for ageing workers.

A comparison with the iron foundry industry, where symptoms were found to affect the entire spinal region and were associated with poor posture and psychosocial stress [5], supports our cluster of neck-trunk-wrist symptoms, though with a greater impact on the wrists (65.9%). In a needle manufacturing industry in Portugal, 54% of complaints were in the thoracic region, 42% in the cervical region and 39% in the shoulders, with a higher incidence among women and older workers [7], mirroring our findings regarding female and age-related susceptibility, thereby validating the demographic vulnerabilities of national manufacturing industries. Furthermore, another example from the metalworking industry revealed predominantly symptoms in the spine and upper limbs due to static postures [8], reinforcing the intercorrelations between neck-shoulder-wrist symptoms and the REBA risk of trunk musculoskeletal disorders. However, our study, which employs this new AI-based computer vision approach, measures dynamic risks, thereby overcoming the limitations of manual observation. Finally, manual work involving porcelain also revealed patterns across multiple anatomical regions, the upper body, spine and elbows, with moderate to high REBA risks [6], and these results are consistent with the findings obtained in the present study. Nevertheless, a key finding of our work is that physical exercise acts as a possible protective factor in preventing WRMSD risks, aligning with the occupational health and operator well-being strategies of Industry 5.0.

Thus, this work validated this new approach and methodology using the NMQ and AI-computational vision for ergonomic assessment in the polishing industry, with gender-age clusters requiring personalised interventions due to their greater vulnerability, namely, female workers and older operators, and thereby contributing to the retention of an ageing workforce in industries.

5. Limitations and Future Work

Some inherent limitations of the case study conducted in a real industrial environment were detected, namely the relatively small sample size (41 polishing operators) and specifically the 27 operators assessed at the ergonomic level, which may restrict the statistical power of the investigation. Thus, it is necessary to point out that, although the sample size is relatively small compared with larger prevalence studies on WRMSDs, the sample reflects the actual number of eligible workers in the polishing department of this industrial sector and, therefore, the results should be interpreted as exploratory and context specific. However, the results could be extrapolated to the entire workforce studied in terms of sociodemographic characteristics and musculoskeletal symptomatology. Although it is not possible to guarantee with 100% certainty that the postures adopted by the operators represent the entire sector, the comparative results between studies have confirmed high ergonomic risks. This may suggest that the studied correlations can be generalised to the polishing operators. However, it should also be noted that the stratification by both age and gender resulted in reduced sample sizes for some sub-groups, limiting the statistical power of these specific comparisons. Given the small size of the subgroups stratified by age and gender, correlation analyses may be subject to an increased risk of false-positive results and should therefore be interpreted cautiously and with exploratory intent. However, this was due to the need to maintain consistency with the methodological strategy of the previous study and to allow a comparison for the validation and assessment of the robustness and reliability of the ergonomic assessment methodology under investigation. Therefore, these findings should be viewed as exploratory and indicative of specific vulnerabilities within this workforce.

Another challenge relates to the use of the TuMeke application, an artificial intelligence-based system for ergonomic assessment, whose computational model and exact details are not fully transparent, limiting independent validation of the results. Moreover, the postural assessments are primarily image-based and may not fully capture the dynamic factors in repetitive work, such as accumulated muscular fatigue and individual variability in susceptibility to injury, restricting a comprehensive interpretation of the risks. Additionally, the potential influence of the observer effect during filming may have altered the natural behaviour of the operators, despite the planned acclimation period.

Regarding the reported symptoms by the operators and the musculoskeletal risks to which they are exposed, small discrepancies were found between the musculoskeletal symptoms reported and the evaluated ergonomic risks in some body regions, suggesting that combining methods or applying additional tools would be desirable to better capture the complexity of ergonomic impact.

Future work should focus on applying this assessment methodology to other industrial contexts, ensuring representative and robust ergonomic evaluations. It is recommended to integrate complementary methods such as sensors and electromyography for direct measurement of muscle activity and analysis of task repetition rates. Future research should realize longitudinal Test–Retest designs and the analysis of effect sizes to further consolidate the reliability of this methodology and to better understand the temporal stability of the risk metrics. Moreover, the calculation of formal statistical reliability indices must be performed, such as Intraclass Correlation Coefficients (ICCs) and Cronbach’s alpha, to transition from the current methodological validation by replication to a comprehensive psychometric validation of the assessment framework. Further proposed research includes exploring assistive technology interventions such as exoskeletons and management strategies aimed at active ageing and promoting sustainable occupational health.

6. Conclusions

This work validated and replicated the methodology applied by Refs. [3,4] in the ergonomic assessment and the prevalence of musculoskeletal symptoms among operators in the metalworking sector, focusing on the task of polishing luxury metal parts. The data confirm the robustness and applicability of the methodology, showing consistent patterns of moderate-to-high ergonomic risk, especially in the regions of the upper limbs and spine, as well as a high prevalence of musculoskeletal symptoms reported by operators, mainly in the neck, shoulders, and wrists. The differences observed between genders and age clusters highlight the greater vulnerability of the female gender and of more experienced workers, that is, those with longer tenure in the company, emphasising the need for targeted ergonomic interventions sensitive to these variables.

Additionally, the correlations identified between health habits, physical activity practices, general health perception, and musculoskeletal risks and symptoms suggest that integrated prevention and occupational management strategies, which include the promotion of health and well-being in the workplace, are crucial for mitigating WRMSDs. The use of innovative tools based on artificial intelligence proved effective for dynamic and non-intrusive assessment, allowing ergonomic risk to be captured without interfering with production processes.

Finally, despite the sample size and the specificity of the investigated industrial context being limiting factors due to their particularity, the work demonstrated validity and reproducibility. Therefore, future research is pointed towards other sectors and environments, with possible application of complementary methodologies, for better understanding and prevention of associated risks. Thus, this work contributed to improvements and innovation in the field of industrial ergonomics, underlining the importance of multidimensional approaches to promote safer, inclusive, and sustainable work environments, compatible with the challenges of an actively ageing workforce.