MODULAR PRODUCT OPTIMIZATION TO ALLEVIATE POVERTY: AN IRRIGATION PUMP CASE STUDY

Transcription

1 Proceedings of the ASME 01 International Design Engineering Technical Conferences & Computers and Information in Engineering Conference IDETC/CIE 01 August 1-15, 01, Chicago, IL, USA DETC MODULAR PRODUCT OPTIMIZATION TO ALLEVIATE POVERTY: AN IRRIGATION PUMP CASE STUDY Charles D. Wood Undergraduate Research Assistant Dept. of Mechanical Engineering Brigham Young University Provo, Utah Patrick K. Lewis PhD Candidate Dept. of Mechanical Engineering Brigham Young University Provo, Utah, Christopher A. Mattson Assistant Professor Dept. of Mechanical Engineering Brigham Young University Provo, Utah, ABSTRACT Modular products have the potential to significantly reduce the financial risks associated with purchasing an income generating product in the developing world. Their modular nature allows a product to adapt to the changing needs of the customer (changing views of affordability due to increase in income potential). In a previous work by the authors, an optimization-based modular product design method was developed and implemented in the design of a modular irrigation pump for poverty alleviation. This paper revisits this modular pump example with the purpose of physically validating the ability of the method to identify theoretical progressively affordable modular products. This paper gives a summary of the method, presents the theoretical pump design, and compares the performance of the theoretical design to a physical prototype of the pump. Based on observations from this comparison, the authors conclude that the method is a feasible approach to engineering-based poverty alleviation 1 Introduction & Background Over the past thirty years there have been notable advances in using engineering as a resource to address the challenges of poverty alleviation efforts [1]. For example there are many engineered products on the market today that have helped more than 1 million people to escape poverty by increasing their income []. These results have served to strengthen the position of those who advocate that the best way to help people escape poverty is to provide them with a way to make more money [3, 4]. Although these income-generating products have had measurable impacts in helping individuals to escape poverty, the large initial investment required to purchase many of them ( -3 months income) has limited their distribution due to the financial risks involved [5 7]. To overcome these financial risks and increase the use and distribution of these products, Lewis et al [8] presented the concept of creating low cost modular products. The advantage of a modular product is the ability to reduce the initial investment required to purchase a product. Furthermore, the income generated through the initial purchase then serves as a means for financing future upgrades that are made affordable through use of the product. As part of the research presented in Lewis et al [8], it was noted that many of the challenges in designing incomegenerating products for those in extreme poverty could be viewed as conflicting design objectives (i.e., minimize cost, maximize income-generation potential, maximize efficiency, etc.). Multiobjective optimization is a well-known, well-accepted means to quantify trade-offs between competing design objectives [9 11], and is a fundamental part of this method. Application of the method was demonstrated through the identification of a theoretical modular irrigation pump for developing countries consisting of a platform and two modules [8]. Due to the challenges in implementing the analysis models of the various configurations of the pump into the optimization Address all correspondence to this author. 1 Copyright 01 by ASME

2 method, the method was expanded to better facilitate more complex design scenarios [1]. The purpose of the present paper is to validate the ability of the method to identify progressively affordable modular products. This is done by revisiting the modular pump example using improved analysis models and design concepts, building a physical prototype of the pump, and comparing the analytical and measured pump performance. By doing this, the authors aim to contribute to engineering-based poverty alleviation efforts by providing both analytical and physical data and help the reader better design for those in poverty. The remainder of this paper is presented as follows: A review of the optimization-based modular-product design method [1] is presented in Section. In Section 3, the optimization results and physical prototype are described in detail. In Section 4 the predicted and measured pump performance is compared. In Section 5, the challenges and potential improvements that are needed in the method are identified. Concluding remarks are provided in Section 6. Modular-Product Optimization Method Summary Fundamental within multiobjective optimization is the identification of a set of non-dominated designs a Pareto frontier [13 15]. Figure 1 illustrates the meaning of non-dominance in the presence of multiple system concepts (bold line) for the minimization of two objectives (µ 1 and µ ). The feasible regions of each system concept are illustrated in the figure as shaded regions. In Mattson and Messac 003 [16], the term s-pareto was adopted to refer to the combined Pareto frontier resulting from a set of disparate concepts. As can be seen in Figure 1, the s-pareto frontier comprises all non-dominated designs because there are no other feasible designs from any other concept that are better in all objectives. A Pareto/s-Pareto frontier, by nature, represents all optimal product candidates. As such, one of the fundamental ideas of the modular-product design method summarized in this section is that the current and future needs of a product can be represented by individual designs along the Pareto/s-Pareto frontier. Building on this assumption, the intended outcome of this method (see Figure ) is that it identifies designs along the Pareto/s-Pareto frontier that facilitate the creation of modules that jump from one location on the frontier to another. These non-dominated designs are selected based on models of present and future needs, and are used to provide target performance values in the design of platforms and modules. This concept is graphically represented for a design situation involving a s-pareto frontier in Figure 1. The designs represented by P (1) P (4) represent the designs that satisfy the present and future needs that define the boundaries of the identified set of anticipated regions of interest. The identified platform and modules are subsequently designed to achieve the desired product performances identified by the designs P (1) P (4). The six steps of the multiobjective optimization design Figure 1. µ Module 3 P (4) Module (4) (4) (3) µ 1l µ1u µ1l µ1u P (3) P () Module 1 (3) () () µ 1l µ1u Concept Model 3 P (1) (1) (1) µ 1l µ1u s-pareto Frontier Interest Regions Concept Model Concept Model 1 Platform Design Graphical representation of the intent/result of the method presented in Figure. Notice that the identified modular product can adapt to designs that are within designer defined regions of interest representing the current and future product needs. method presented in Figure are: Step A: Evaluate and characterize the effect of each design variable on the multiobjective design space characterize the Pareto frontier of each design concept. Step B: Indicate predicted changes in customer needs by specifying anticipated regions of interest within the multiobjective design space as a series of upper and lower objective limits over time (See Figure 1). Step C: Identify platform design variables (x p ) that result in minimal losses in objective space performance within the anticipated regions of interest. Step D: Use Pareto filtering methods to identify the s-pareto frontier within each region of interest. Step E: Implement an optimization routine to identify the target s-pareto-optimal designs within each anticipated region of interest that are best suited to collectively facilitate traversing the s-pareto frontier. The formulation of this optimization routine can be found in Lewis et al [1]. Step F: Identify the module designs by evaluating a constrained module-optimization design routine. This is done by: (i) Selecting a modular architecture type [17 1], (ii) Identifying the product platform design and module interfaces [, 3], (iii) Determining the desired number of modules and modular progression, and (iv) Identifying and calculating the values of module design variables through constrained module optimization. As was mentioned in Section 1, the development of this method was divided into two phases. The first phase of method developments focused on traversing a single Pareto frontier, µ 1 Copyright 01 by ASME

3 A B C D Characterize the Multiobjective Design Space Define Anticipated Regions of Interest Select Platform Variables No Platform Variables Selected? Yes Assemble the s-pareto Frontier Within Each Region of Interest (L/s) Hip Pump Plus Super Sales Price ($) Figure. Flow chart describing the six-step optimization-based modular product design method. E F Select the Optimal Design Within Each Region of Interest Develop Modules That Move From One Region of Interest to Another Select a Modular Architecture Type Determine the Desired Number of Modules and the Modular Progression Identify the Product Platform Design and Module Interfaces Identify and Calculate the Values of Module Design Variables Figure 3. Graphical comparison of three non-modular water pumps that are currently sold on the market. The horizontal axis represents the sales price (S) in US dollars, and the vertical axis represents the potential water flow rate (Q) in liters per second [8]. In short, the motivation for a modular pump comes from the need to better design for those in poverty, that is, reduce the financial risks involved with purchasing traditional irrigation pumps [5 7] while still providing the needed pump performance that will increase the purchaser s income. For example, a modular irrigation pump can be purchased in a series of upgrades. The initial investment of the first platform is smaller, yet will yield a modest increase in income. This income can then be used to finance the purchase of the second module with said upgrade increasing the pump performance and subsequently its income generating potential. Building on this motivation, the next subsections follow the progression of a modular pump from the theoretical to physical development. while the second phase extended the method to s-pareto design situations. For complete presentations of these method developments, see Lewis et al [4]. 3 Modular Irrigation Pump Design The purpose of this section is to describe the theoretical and physical prototype pump design that resulted from the method described above. As described in Lewis et al [8], drive for the development of this modular pump is best illustrated by considering three non-modular irrigation pumps that are currently sold on the market today and represented in Figure 3 in terms of the flow rate (with units L/s) and sales price (S in US dollars). The products represented in Figure 3 can be viewed as satisfying a range of what is currently considered affordable, but none of these products are capable of expanding as an individual s view of affordability changes due to increases in income potential (i.e. a Hip Pump cannot become a Super ). 3.1 Theoretical Pump Design As described in Section 1, analytical models of the fluid (derived from the Energy Equation of the First Law of Thermodynamics) and financial (sum of the predicted cost of all pump components) aspects of the modular irrigation pump were developed and implemented in the method presented in Section. The developed models of the fluid, income-generation potential, and financial aspects of an irrigation pump implemented the basic architecture illustrated in Figure 4 to predict the behavior of a pump design. A complete list of the assumptions made in the development of the disparate financial and fluid models, along with detailed descriptions of these models, are presented in Lewis et al [5]. Descriptions of the variables and parameters shown in Figure 4 are as follows: l o is the distance (m) from the pivot to the operator; l c is the distance (m) from the pivot to the pump cylinder; d p is the inner diameter (m) of the inlet/outlet pipe; d c is the inner diameter (m) of the piston cylinder; h c is the distance 3 Copyright 01 by ASME

4 l o F l,f h 1.8 Super l c l s h c z z out in l p,out d c d p Figure 4. Graphical illustration of the basic pump architecture and loading l p,in conditions used in the development of the analytical pump models. Table 1. Variable and objective values of the platform pump design. Variables Objectives d c (m) l o (m) l c (m) n c (m) Q (L/s) S ($) (L/s) Sales Price ($) Figure 5. Hip Pump Plus Platform Design Module 1 Module Graphical comparison of the predicted modular pump iterations and the three benchmark non-modular water pumps. Table. Variable and objective values of the modular pump iterations obtained by adding the i-th module design. (m) traveled by the cylinder piston head; l s is the length (m) of the operator stroke; l p,in is the length (m) of the inlet pipe; l p,out is the length (m) of the outlet pipe; z in the vertical distance (m) from the pump to the water source; z out is the vertical distance (m) from the pump to the pipe outlet; and F is the force applied by the operator (N) during hand and leg operation of the pump. The modular pump concept that resulted from implementing the presented method is a hand-operated irrigation pump that transforms into a two-cylinder treadle irrigation pump through the addition of two modules. These modules are characterized by the following configuration descriptions: (1) Hand actuated with single cylinder, () Foot actuated with a single cylinder, and (3) Foot actuated with two cylinders. The variable values of the platform and module designs are presented in Tables 1 and respectively, where l t is the length of the i-th treadle extension (m) and ˆn c is the number of cylinders added by the i-th module. A representation of the predicted Q and S for the various configurations of the pump are provided in Figure 5, along with the target pump designs from Figure Physical Hardware Using the optimization results presented in Tables 1 and, a physical prototype of the pump was created. The purpose of this subsection is to present the physical modular pump design and compare the actual dimensions and sales price (in US dollars) of the prototype to the optimization results from Section 3.1. Variables Objectives i d c (m) l t (m) l c (m) ˆn c (m) Q (L/s) S m ($) It should be noted that the optimization results from the previous subsection only provided the dimensions for key attributes of the modular pump (cylinder location, length of treadles/treadle extensions, etc.). This is because a conceptual model (not a detailed model) was used to optimize the design. In order to create a physical prototype, we developed a detailed concept of the pump design. This included defining (i) the shape of the pump support frame; (ii) the material and cross-section of the pump frame members, handle, and treadles; and (iii) the module interfaces for the handle, the single/double cylinder configurations of the pump pistons, and the frame/treadle extensions. Using the optimization results from the previous section, the physical prototype of the detailed pump concept was then built to match the identified key dimensions (see Tables 3 and 4). Figure 6 shows the physical prototype that was developed. In order to facilitate the discussion of the theoretical and physical prototype pump performance in Section 4, the measurements of the key dimensions reported in Tables 1 and, along with the measured values of the operator stroke length (l s ) of 4 Copyright 01 by ASME

5 (a) (b) (c) Figure 6. Three configurations of the modular irrigation pump. The first configuration (a) is a hand operated single-cylinder pump indicated in blue. The second configuration (b) is a foot operated single-cylinder pump with the upgrade identified in yellow. The third configuration (c) is a foot operated dual-cylinder pump with the upgrade indicated in red. Table 3. Table 4. Variable values of the prototype platform pump design. d c (m) l o (m) l c (m) n c (m) l s (m) Variable values of the prototype modular pump iterations obtained by adding the i-th module design. i d c (m) l t (m) l c (m) ˆn c (m) l s (m) each pump iteration are provided in Tables 3 and 4. 4 Physical & Theoretical Pump Performance Comparison In order to compare the physical prototype to the theoretical pump performance, each iteration of the pump was tested for values of z in and z out ranging from meters (0 8.5 meters total change between the inlet and outlet of the piping). In all, the hand operated and single cylinder treadle pump iterations of the pump were tested eighteen times, and the dual cylinder treadle pump iteration was tested thirteen times. The results of these tests in terms of the measured flow rate of the pump are presented in Figure 7. Also included in this figure is the predicted flow rate of the pump for the specified values of z in and z out, along with the values of d c, l o, l c, n c, l s, and l t presented in Tables 3 and 4. As indicated in Figure 7, the measured flow rate of the physical prototype is represented by the solid lines, and the predicted flow rate is represented by the dashed lines. From the results in Figure 8(a), it can be seen that the platform pump design recorded a higher flow rate then that predicted by the analytical fluid model. Based on these results, the average difference between the measured and predicted flow rate for this iteration of the pump is 0.06 L/s. Similarly, for the plots provided in Figures 8(b) and (c), the average difference between the measured and predicted flow rate for the single and dual cylinder treadle pump iterations are 0.03 and 0.08 L/s respectively. It is noted that there is a difference in the maximum flow rate values shown in Figure 7 compared to those shown in Figure 5. This is due to the setup required to perform the tests recorded in Figure 7. One example of changes in the setup that were not included in the execution of the optimization routines are the 50 meter hoses used to draw the water from 4 meters and push it up to 4 meters. As a result, in the 100 meters of hose, there are additional losses in the system that were not considered in the previous optimizations. However, from the discussion of the differences between the predicted and measured flow rate of the pump, the models can still be used in the future to improve the optimization results, and subsequent pump performance to provide an overall better design for people living in poverty. From the plot axis of Figure 5, it is observed that in designing for those in poverty, the selected optimization objectives were to minimize the sales price and maximize the flow rate of the pump. As such, to complete the comparison of the pump performance the measured and predicted sales price of the pump platform and modules are presented in Table 5. From this table 5 Copyright 01 by ASME

6 5 Single Cylinder Hand Operated Pump 0.55 Single Cylinder Treadle Pump Measured (L/s) 5 (L/s) Measured Predicted Predicted z +z (m) in out (a) z +z (m) in out (b) 1.4 Dual Cylinder Treadle Pump 1. 1 (L/s) Predicted Measured Figure z in +z out (m) (c) Comparison of the measured and predicted flow rate of the different iterations of the modular pump for values of z in and z out ranging from meters (0 8.5 meters total change between the inlet and outlet of the piping). it can be seen that the measured price of the prototype, adjusted for assembly production, is able to follow the same approximate trend predicted by the optimization. This result, along with the correlation between the predicted and measured flow rate of the pump, shows that the concept of a progressively affordable pump and the method used to obtain this design are valid and feasible. 5 Observations & Future Work Recall that the method described in Section was used to determine the values of the key design variables. When developing concepts of how to make the pump with the desired modules, several challenges were encountered. In the initial concepts that were considered, the cylinders were aligned below the treadles with the user unsafely positioned higher off the ground (see Figure 8). In order to keep the user as close to the ground as possible, the design was changed and the cylinders were repositioned above the treadles as shown in Figure 6. An unanticipated result of repositioning the cylinders with the pistons facing down was a loss in the efficiency of the piston seal that resulted in leaking. Unfortunately, this was not observed as a problem until tests were run trying to pull/push water more than two meters below/above 6 Copyright 01 by ASME

7 (a) (b) (c) Figure 8. CAD models of initial concepts of the modular pump show the cylinders aligned below the treadles. The iteration shown includes the platform fig:platformresults, module one fig:mod1results and module two fig:modresults. Table 5. Comparison of the measured and predicted sales price (in US dollars) of the pump platform (i = 0) and modules (i = 1,). i Measured Price ($) Predicted Price ($) the pump. Although the use of modularity is key in this paper, there was another unexpected consequence of using this design approach. Modularity necessitates the design of many different individual components where normally one single component could be used. Although this allows for reconfiguration to facilitate the purchase of incremental upgrades by reducing the initial financial risk, the increased number of joints gives way to increased likelihood of structural failures at the module interfaces. The method presented in Section focuses on identifying designs that target desired values for the design objectives that represent the goals and preferences of the target customers. One observed limitation of the method is that it does not currently include a system of accounting for the economic, social, environmental, technological and political factors that determine the success of a product upon implementation in the developing world. Future work to be done in this area will focus on the identification of a metric that can measure, and potentially predict, how successfully a product accounts for the complexities of these social factors. In addition, future work will also demonstrate the robustness of the presented optimization-based modular product design method to many applications, including those outside of poverty alleviation efforts. Finally, due to the inherent physical exertion that is required to operate the pump, the physical endurance of the person operating the pump is an important issue. In the current optimization routine described in Section 3.1 a constraint that limits the minimum time to complete a single pump stroke to 0.5 seconds was intended to address this issue. However, during the pump testing described in the previous section, it became clear that the physical endurance of the pump operator was not adequately considered and represented in the implemented optimization routine. Although the operator did not have a stroke time that was less then 0.5 seconds, the required stroke time to achieve the flow rate data and results shown in Figure 7 would not be physically sustainable for an extended period of time. As such, future work will also include indentifying constraints that ensure the endurance of the pump operator will be adequately represented and accounted for in making design decisions. 6 CONCLUDING REMARKS By using engineering techniques to find solutions to further poverty alleviation efforts, there have been many advancements. Through the modular pump example presented in this paper, it can be observed that modular product design is a feasible approach to significantly reducing the initial financial risks associated with purchasing income generating products. This paper summarized a method capable of balancing the income generation potential of a product and the investment cost through optimization-based modular product design, thereby providing a better design for developing nations. In this work, a prototype modular irrigation pump with three configurations was presented, and tested to validate the theoretical results of the presented method. In Section 4 the measured and predicted flow rates and sales price for all three pump configurations are provided, and compared. In hopes of helping the reader better design for people in poverty, observations of the challenges of transitioning from the theoretical design to physi- 7 Copyright 01 by ASME

8 cal prototype and the impacts that these challenges can have on the performance of a product are also provided. Future work related to this research should seek to address these challenges within the method. Additional work should also look at developing methods for identifying the anticipated regions of interest. This is especially important when developing income generating products due to the inherent impact these regions have on the modular products identified, and the ability these products have to reduce the perceived financial risks. ACKNOWLEDGMENT We would like to recognize the National Science Foundation Grant CMMI for funding this research. REFERENCES [1] The World Bank, 010. World development indicators 010. Tech. rep. 1 [] Polak, P., 005. The big potential of small farms. Scientific American, [3] KickStart, 009. Our 5 step process, May. 1 [4] Enterprises, I. D., 007. Enabling Prosperity: 007 Annual Report. 1 [5] Fisher, M., 006. Income is development. Innovations Journal, Winter 006, pp , 3 [6] Johnson, N. G., Hallam, A., Bryden, M., and Conway, S., 006. Sustainable and market-based analysis of cooking technologies in developing countries. In Proceeding of the ASME International Mechanical Engineering Congress and Exposition, no. IMECE , 3 [7] World Resources Institute and International Finance Corporation, 007. The next 4 billion: Market size and business strategy at the base of the pyramid. Tech. rep., World Bank, June. 1, 3 [8] Lewis, P. K., Murray, V. R., and Mattson, C. A., 010. An engineering design strategy for reconfigurable products that support poverty alleviation. In the ASME 010 International Design Engineering Technical Conferences & Computers and Information in Engineering Conference, no. DETC , 3 [9] Kasprzak, E. M., and Lewis, K. E., 000. An approach to facilitate decision tradeoffs in pareto solution sets. Engineering Valuation and Cost Analysis, 3, pp [10] Messac, A., Physical programming: Effective optimization for computational design. AIAA Journal, 34(1), pp [11] Frischknecht, B. D., Peters, D. L., and Papalambros, P. Y., 011. Pareto set analysis: local measures of objective coupling in multiobjective design optimization. Structural and Multidisciplinary Optimization, 43(5), pp DOI: /s [1] Lewis, P. K., and Mattson, C. A., 011. A method for developing systems that traverse the pareto frontiers of multiple system concepts through modularity. Structural and Multidisciplinary Optimization, Accepted in October. [13] Messac, A., and Mattson, C. A., 004. Normal constraint metthod with guarentee of even representation of complete pareto frontier. AIAA Journal, 4(10), pp [14] Todoroki, A., and Sekishiro, M., 008. Modified efficient global optimization for a hat-stiffened composite panel with buckling constraint. AIAA JOURNAL, 46(9), September, pp DOI: 1514/ [15] Gurnani, A. P., and Lewis, K., 008. Using bounded rationality to improve decentralized design. AIAA JOURNAL, 46(1), December, pp DOI: 1514/ [16] Mattson, C. A., and Messac, A., 003. Concept selection using s-pareto frontiers. AIAA Journal, 41(6), pp [17] Yang, T. G., Beiter, K. A., and Ishii, K., 004. Product platform development: An approach for products in the conceptual stages of design. In In 004 ASME International Mechanical Engineering Congress and RDD Expo, no. IMECE [18] Ulrich, K., and Eppinger, S., 004. Product Design and Development, 4th ed. McGraw-Hill. [19] Oyebode, A., 004. Modularity and quality. In Proceedings From the nd Seminar on Development of Modular Products, G. Erixon and P. Kenger, eds., pp [0] Padamat, M., 004. Methods for modularisation. In Proceedings From the nd Seminar on Development of Modular Products, G. Erixon and P. Kenger, eds., pp [1] Strong, M. B., Magleby, S. P., and Parkinson, A. R., 003. A classification method to compare modular product concepts. In the ASME 003 Design Engineering Technical Conferences. [] Tseng, M. M., Jiao, J., and Merchant, M. E., Design for mass customization. CIRP Annals - Manufacturing Technology, 45(1), pp [3] Simpson, T. W., 004. Product platform design and optimization: Status and promise. Artificial Intelligence for Engineering Design, Analysis and Manufacturing, 18(1), pp [4] Lewis, P. K., Murray, V. R., and Mattson, C. A., 011. A design optimization strategy for creating devices that traverse the Pareto frontier over time. Structural and Multidisciplinary Optimization, 43(), February, pp DOI /s [5] Lewis, P. K., Murray, V. R., and Mattson, C. A., 010. Accounting for changing consumer needs with s-pareto frontiers. In the 13th AIAA/ISSMO Multidisciplinary Analysis Optimization Conference, no. AIAA Copyright 01 by ASME