wheel hoe 2kx final report · 2018. 9. 6. · ! 3!...
TRANSCRIPT
1
Wheel Hoe 2KX: Design for the 21st Century Woman IE 547: Designing for Human Variability
Kelly Sprehn Yang, Hui
Executive Summary Given the increase in number of women involved with farming and gardening activities and the lack of women-‐optimized tools in the market, this study examines the design of a wheel hoe for women. The following table demonstrates the static design and adjustable design measurements suggested by the study.
Mean Range Handle Diameter 30.4 17.7 – 42.9 Handle Width 382 352 – 410 Height 831 723 – 939
A prototype was tested by 12 females. This preference data was correlated to various anthropometric measures and expanded in a hybrid model using both NHANES and ANSUR databases. Recommendations take into account the accommodation of the central 95% of the target population. While the study is limited in the number of participants, the tested posture, and the subsequent model correlation, it is believed that this study guides preliminary development of a wheel hoe designed specifically for women. Introduction Necessity According to the U.S. Census numbers, the number of women who own and operate farms has increased 46% from 1997 to 2007 (U.S. Statistics on Women and Minorities on Farms and in Rural Areas, 1997, 2002, 2007). That trend is continuing as the economy encourages more people to grow their own food and trends of purchasing organic food continue to rise. The popularity of the farmer’s market, a gathering of growers and food-‐conscious buyers, has increased significantly in the past few years. The growth of small farmers, market growers, and hobbyists leads to an emerging population needing tools to help in their tasks. A significant risk to this population of women is the differences in body types from that of men. In many design scenarios, anthropometric differences would not have a significant impact on the use or risk of the product. However, according to a study by Hsaio et al. (2002), women in agriculture differ significantly in 11 of 14 body measures (Table 1). Green Heron Tools, a women-‐owned, women-‐focused gardening and farming company, conducted preliminary research in this area. Their anecdotal evidence suggests that many women find garden tools “too heavy”, “too long”, “too high”, “not well balanced”, among others (Green Heron Tools, 2010).
2
Table 1: Anthropometric Measurements of Agricultural Workers in the U.S.
Sitting Height* Waist Circumference* Upper Arm Length* Buttocks Circumference* Upper Leg Length* Thigh Circumference* Biacromial Breadth* Wrist Breadth* Biiliac Breadth* Arm Circumference* Elbow Breadth* Weight Stature BMI
* Indicates that this measure differs significantly between men and women. Demographics Based on this necessity, the target demographics are derived from the U.S. Census Bureau report of Women and Minorities on Farms and in Rural Areas. In combination with this study, Hsaio, et al. (2002) provides more insight into the anthropometric measurements of this population (Table 2).
Table 2: Sample of Anthropometric Measurements of U.S. Female Agriculture Workers
Variable Mean 95% CI
Stature (mm) 1592 1578 – 1607
Weight (kg) 68.7 65.9 – 71.6
BMI 27.2 26.1 – 28.2
Biacromial Breadth (cm) 36.2 35.9 – 36.6 Design Process Background The variability in anthropometry indicates the adjustability and sizes of artifacts to accommodate the target users. In the design of artifacts using spatial dimensions of the user population, there are several general approaches to achieving users’ accommodation: manikins, population models, and the hybrid model approach. A boundary manikin refers to a body measurement according to the limit of acceptability. The manikin approach typically uses two-‐ or three-‐dimensional to represent the human body size and shape. Drillis et al. (1966) highlighted the impact of this approach by claiming a set of proportionality constants calculated for a sample population. This approach intends to quantify the anthropometric variability expected within the target population of users. It is often used in the absence of actual data for the target user population. However, the main limitation of this approach is that it is only works in extremely constrained cases, because only boundary anthropometries are utilized and that might produce misleading results.
3
The population model approach creates models by experimental data from the representative sample population performing a task that is related to the dimension under study (Roe, 1993). It models the specific target measurement and thus greatly improves the results compared with the manikins approach. However, the only factor in this method is the body dimension. It is known that two users with similar body dimensions might have different preference on the target measurements. Without considering the preference of the sample population, using this approach can produce inaccurate results. Hybrids models overcome the limitations of the above two approaches by importing a stochastic component based on the residual variance in the regression analysis (Garneau et al, 2009). The anthropometry-‐driven preference model contains the variability of body and also includes the remaining variability, thus provides more accuracy to the predictions (Nadadur et al., 2008). This approach has been shown to integrate consideration of variability that is not correlated with the predictors in several applications such as predicting automobile driving posture (Reed et al., 2002), optimizing vehicle occupant packaging (Parkinson et al., 2006). Parkinson et al. (2010) present a statistical method for applying the available anthropometric data to estimate distributions of the anthropometric data for a target population. The virtual population that generated by the anthropometric data via this method can be used to represent the target user population. The hybrid method and use of virtual populations will be utilized in this project to optimize the wheel hoe, as described below, for a target user population. The virtual population in this project will be obtained by random sampling in the data from the representative databases NHANES and ANSUR. Prototype By developing tools for women, the increased risks of back injuries, work-‐related musculoskeletal injuries, and discomfort can be mitigated. While this call reaches every tool in the garden shed, this study will focus on the wheel hoe (Figure 1). This implement was chosen for the availability of parts, the ability to modify certain areas of the tool, and its use by women in the gardening and farming scenarios mentioned earlier. A 3-‐D CAD model was created to generate a better understanding of the parts to be modified (Figure 1 – 4). This model will also serve for future testing, design ideas, and prototype generation.
4
Figure 1: Wheel Hoe Figure 2: 3D Model of Wheel Hoe
width
Figure 3: Adjustability of Wheel Hoe Figure 4: Adjustability of Wheel Hoe
The purchase of a Whiz Bang Wheel Hoe kit provided the basic parts for a testable prototype. These rough metal parts were assembled with adjustable shafts and handles along with different handle diameters (Figure 5, Table 3).
Table 3: Dowel Diameters (mm)
Dowel Number Dowel Diameter 1 36.75 2 30.63 3 24.50 4 18.38 5 15.33 6 12.25
Figure 5: Prototype of Wheel Hoe
5
Testing In order to find the preference of use for various people, tests were run. First, anthropometric measurements were collected. These included Stature (mm), Weight (kg), Wrist-‐Hand Length (Figure 6), Biacromial Breadth (Number 10, Figure 7), and Forearm-‐Forearm Width (Number 53, Figure 7). After this, subjects were asked to choose between the six handle diameter sizes. They were allowed to try different sizes without being attached to the tool in order to find a grip that was most comfortable. This handle was then attached to the wheel hoe and subjects were asked to identify the most comfortable position for the bottom and handle angles (Figures 8 – 11). These were moved by the experimenters to avoid injury of the participants. Overall, 12 participants’ data were used in this study (Appendix A).
Figure 6: Hand Length(mpt,2009) Figure 7: Anthropometric Measurements(ANSUR)
Figures 8-‐11: Adjustability shown on Prototype (top) and CAD models (bottom)
01
2
43
Handle Angle
-165
55
45
351
234
Bottom Angle
6
Modeling In order to find the most appropriate design recommendations for the wheel hoe, grip diameter, height, and handle width are the factors that need to be considered. The anthropometry of the 12 collected data points was compared to the anthropometric measurements of female agricultural workers as described by Hsaio, et al. (2002). This comparison can be seen in Table 4.
Table 4: Comparison of Target Population and Sample Data
Sample Population
Variable Mean 95% CI Mean 95% CI
Age 54 26
Stature (mm) 1641 1603 -‐ 1680 1592 1578 – 1607
Weight (kg) 65.2 57.0 – 73.3 68.7 65.9 – 71.6
BMI 24.1 21.4 – 26.9 27.2 26.1 – 28.2
Biacromial Breadth (cm) 37.2 35.6 – 38.7 36.2 35.9 – 36.6
By examination, the sample does not accurately represent the population. There are several ways to account for these differences to achieve an accurate prediction model. The hybrid modeling approach allows researchers to use a small sample data, find correlation between factors, and apply that correlation and variation to a larger sample size, typically generated from a population database such as ANSUR (Army ANthropometric SURvey) or NHANES (National Health and Nutrition Examination Survey). In generating this hybrid model, there were several options. To find the best representative model, the sample itself was considered. ANSUR and NHANES databases were also compared. ANSUR provides very detailed and precise measurements over many parts of the human body. Specific measurements can be used, however, the population is limited to the Army personnel of 1987-‐1988. Comparing this to the target population, the ANSUR population would match well with a full-‐time agricultural worker, but not represent the small gardener or hobbyist. NHANES provides a better representative sample due to the demographics of the study. However, NHANES is limited to providing measures only about gender, age, stature, and weight. In order to predict grip diameter, hand length, stature, and the natural log of the body mass index were correlated to grip diameter preference. The prediction of wheel hoe height, as calculated from the chosen bottom angle and shaft length, was correlated to stature and the natural logarithm of BMI. To predict handle width, biacromial breadth, forearm-‐forearm width, natural logarithm of BMI, and stature were individually correlated to preference. These correlations were measured through R2 and served to find the best predictor for each factor. These values can be seen in Table 5.
7
Table 5: R2 Values
R2 Values
Biacromial
Breadth
Forearm
Breadth Hand Length Stature Ln(BMI)
Grip Diameter 0.28 0.302 0.1534
Handle Width 0.2059 0.0056 0.0217 0.1392
Height 0.1003 0.0877
Due to the low values of these correlations, multivariate regressions were compared to find the best possible preference predictor. These will be shown and discussed later. Once a prediction equation was chosen, the equation plus a residual variance measure was applied to a sample population derived from the NHANES database. At a target accommodation level of 95% of the target population, recommendations were generated and compared to measurements of current and available wheel hoe models. Results The experimental data was collected from 12 subjects using a prototype of the wheel hoe. Preference models for each target measure are created by the regression equation after comparison with the added preference term.
Y = a X + b + N (0, RMSE) The N (0, RMSE) comes from the random sampling from a normal distribution with standard deviation equal to the root mean squared error (RMSE) of the corresponding regression. Grip Diameter Due to the similar R2 values indicating the correlation between grip diameter and anthropometric measures, combinations of these measures were analyzed for the best fit.
Handle diameter = 5.05 * Hand Length – 56.38, R2 = 0.2799 (1) Handle diameter = -‐13.01 * ln (BMI) + 72.42, R2 = 0.1534 (2) Handle diameter = 0.051 * Stature – 52.57, R2 = 0.3022 (3)
The relatively large R2 value in regression model (3) reflects that the stature and handle diameter have relatively strong correlation compared with the other two models. Thus, the handle diameter is modeled only in terms of stature.
Handle diameter = 0.051 * Stature – 52.57 + N (0, 5.54) (4)
8
1000 female statures were then sampled using weights to represent the relative presence of a particular data point over eight years of the NHANES database (1999-‐2007). This constitutes the virtual population for the handle diameter. This equation (4) is applied to each member of the virtual population. Using the linear regression modal developed from the preference study, a plot of handle diameter versus stature of the virtual population is given in the following figures (12 – 13).
Figure 12: Sample Data and Regression Figure 13: Regression applied to NHANES data
The following table summarizes percentile data from the result (Table 6). Dimensions are given in millimeters.
Table 6: Handle Diameter in Virtual Population
0% 2.5% 5% 10% 25% 50% 75% 90% 95% 97.5% 100% 9.9 17.7 19.7 22.1 26.5 30.4 34.9 38.9 41.6 42.9 56.5
For a desired level of accommodation 95%, the central 95% percentile values of the virtual population are selected to determine the minimum handle diameter of 17.7mm and the maximum diameter of 42.9mm. Users above and below these cutoffs would be disaccommodated. Wheel Hoe Height The same method is used to determine the accommodation of the height of the wheel hoe. The measurement are obtained by the following equation:
Wheel Hoe Height = length * sin (bottom angle) + bottom height (5) The length is the distance from the bottom of the shaft of the control pole to the upper terminal endpoint of the handle. The bottom height refers to the height of the fixed rotating shaft of the control pole in bottom disc with respect to the ground, which is 35mm in the experiment. The comparison of several outcomes of
9
preference models for handle diameters against three measures of anthropometry are as follows.
Height = 75.382 * ln(BMI) + 586.82, R2 = 0.0877 (6)
Height = 0.2254 * Stature + 455.6, R2 = 0.1003 (7)
Height = 0.2166 * Stature + 72.00 * ln(BMI) + 242.00, R2 = 0.1803 (8)
Compared with the low R2 values obtained from the regression models using single predictor, the relatively higher R2 value in the multi-‐regression model (8) reflects that the height of Wheel Hoe the stature and natural logarithm of BMI has relatively strong correlation compared with the other two models. Thus, the handle diameter is modeled in terms of stature and natural logarithm of BMI. These two predictors are available in the NHANES data, which is preferred to ANSUR due to the better representation of the target population. The preference model is given as follows.
Height = 0.2166 * Stature + 72.00 * ln(BMI) + 242.00 + N(0,48.5) (9)
Equation (9) is applied to the same virtual population from NHANES that contains the randomly sampled 1000 female data in stature and Ln(BMI). The following table summarizes percentile data from the result (Table 7). Dimensions are given in millimeters.
Table 7: Wheel Hoe Height
0% 2.5% 5% 10% 25% 50% 75% 90% 95% 97.5% 100% 608 723 743 763 795 831 870 900 922 939 1034
For a desired level of accommodation 95%, the central 95% percentile values of the virtual population are selected to determine the minimum height of 723 mm and the maximum height of 939 mm. Handle Width The same method is used to determine the accommodation of the handle width. The comparison of several outcomes of preference regression models for handle width against three measures of anthropometry are as follows. Biacromial Breadth is abbreviated as BAB. Handle Width = 101.45 * ln(BMI) + 69.476 , R2 = 0.1392 (10)
Handle Width = 0.8467* BAB + 75.981 , R2 = 0.2059 (11)
Handle Width = 57.95 * ln(BMI) + 0.669 * BAB – 41.58 , R2 = 0.2423 (12)
10
Compared with the linear model (10) and (11), there is no significant increase for R2 value in the multi-‐regression model. In addition, the representative data of BMI and Biacromial Breadth for U.S. female are located in ANSUR and NHANES respectively. The multi-‐regression model increases the effect of the error compared to the linear regression model. Thus, the handle width is modeled in terms of biacromial breadth as the following equation.
Handle Width = 0.8467* BAB + 75.981 + N(0,4.84) (13) 500 female statures are then sampled at random from ANSUR women data and constitute the virtual population for the handle width. This equation (13) is applied to each member of the virtual population. Using the model developed from the preference study, a plot of handle width versus biacromial breath of the virtual population is given in the following figures. (Figures 14 – 15).
Figure 14: Sample Data Regression Figure 15: Regression applied to ANSUR Sample Data
The following table summarizes percentile data from the result (Table 8). Dimensions are given in millimeters.
Table 8: Handle Width
0% 2.5% 5% 10% 25% 50% 75% 90% 95% 97.5% 100% 332 352 357 364 373 382 391 400 406 410 425
For a desired level of accommodation 95%, the central 95% percentile values of the virtual population are selected to determine the minimum handle width of 352 mm and the maximum handle width of 410 mm.
250
300
350
400
450
500
300 350 400 450
11
Discussion Design Recommendations Based on the previous analysis, the recommendations for Handle Diameter, Handle Width, and Wheel Hoe Height are given in Table 9. All dimensions are given in millimeters.
Table 9: Recommendations
Mean Range
Handle Diameter 30.4 17.7 – 42.9
Handle Width 382 352 – 410
Height 831 723 – 939
Generally, wheel hoes are designed to be static, without much adjustability. The mean measurements would provide guidance to development of a static tool but would reduce accommodation to a just-‐noticeable-‐difference population. While this concept was not tested, it is posited that while incorporating adjustability to the design, a static design to the mean would increase the comfort level for a female user population. Market Comparison Given in Table 10 is a brief comparison of market offerings of wheel hoes. When the dimensions were listed in the product description, they have been put into the table. When compared to the mean and ranges recommended by the previous study, it is obvious that the handle length, which contributes to the height, is much longer for market offerings. Handle width is also larger. The weight was not tested in the previous experiment due to the extra components added to the prototype to create the adjustability. Cost, also was not considered, but was included in the table to guide future designs.
Table 10: Market Comparison
Tool Name Handle Length (mm)
Handle Width (mm)
Weight (kg)
Cost
1. Deluxe Hoss Wheel Garden Hoe
1384 419 8.85 $250.00
2. Glaser Professional Wheel Hoe
1448 Unknown but set width
11.3 $350.00
Limitations
These recommendations do have their limitations. First, the correlation of the preferences does not correspond well with anthropometry. This could be due to
12
a number of reasons, not least of which is the small sample size used when obtaining the preference measurements. The age distribution of the 12 people sampled did not match up with that of either the population or of the NHANES data. This would be an issue because the more experienced and developed the subject; their preferences may correlate better and provide better and different predictions of the appropriate tool size. Another factor was the testing not being representative of the use of the wheel hoe. In typical farming and gardening conditions, the wheel hoe is meant to be used in soil and at a walking pace. This situation was not replicated as the subjects adjusted the tool to their preference. Given these shortcomings, future studies should be completed with larger, more representative samples, and in a similar situation to which the wheel hoe can be applied. Conclusion The increase in women participating in farming and gardening activities provides the motivation to develop tools specifically for women. This opens up a niche market and if used, will reduce the risk for injury. Through the use of hybrid models, this study has provided guidelines for the development of a wheel hoe fit for women’s anthropometry. While these measurements offer a good starting point and open up discussion of developing tools for women, the limitations of this study imbue a sense of caution when implementing these recommendations. Future studies should focus on accurate population representation and consistent posture testing when examining the preference of using the tool. Combined with the methodology and preliminary results presented here, these recommendations can make for a very strong case to develop tools appropriate for the comfortable use by women.
13
References Drillis, R., and Contini, R., (1966), Body Segment Parameters, Office of Vocational Rehabilitation Engineering & Science, New York, NY. Garneau, C. and Parkinson, M. (2009). Including preference in anthropometry-‐driven models for design. ASME Journal of Mechanical Design, 131(10):6. Hsiao, H., Long, D., and Snyder, K. (2002). “Anthropometric differences among occupational groups,” Ergonomics, 45(2), pp. 136-‐152. Nadadur, G., and Parkinson, M. B., (2008), “Extrapolation of Anthropometric Measures to New Populations,” SAE International Journal of Passenger Cars Electronic Systems, 1(1), pp. 567–573. Parkinson, M. B. and Reed, M. P. (2010) “Creating virtual user populations by analysis of anthropometric data”, International Journal of Industrial Ergonomics, 40(1), pp. 106-‐111 Parkinson, M., and Reed, M., (2006), “Optimizing Vehicle Occupant Packaging,” SAE Transactions: Journal of Passenger Cars–Mechanical Systems, 115(6), pp. 890–901. Parkinson, M. B. and Reed, M. P. (2009). “Creating virtual user populations by analysis of anthropometric data.” International Journal of Industrial Ergonomics, preprint submitted 2009. Reed, M. P., Manary, M. A., Flannagan, C. A. C., and Schneider, L. W., (2002), “A Statistical Method for Predicting Automobile Driving Posture,” Human Factors, 44(4), pp. 557–568. Roe, R. (1993). “Occupant packaging,” Automotive Ergonomics, pp. 11–42. Taylor & Francis, London, UK