To Editor Thank you for the valuable comments

To Editor Thank you for the valuable comments on this article. We carefully revised the paper according to the comments. The main corrections are as follows: 1. The title, abstract and introduction were revised to clearly point our novelty and contributions of the proposed method. 2. All pictures in the text have been changed to higher resolution pictures to increase the readability of the image. 3. Tables describing the variables are combined into one table to help easy understand variables and pictures. 4. To confirm the cleaning performance, a simple additional test was performed, and the contents were added in the answer sheet. 5. In order to increase the readability and clarity, the text was corrected grammatically by the native after the contents were revised. Please refer the attached revision summary and revised draft for details. The revised contents are highlighted by colors as follows: Review #1 (revision is denoted as Blue highlights in the draft) Review #2 (revision is denoted as Green highlights in the draft)

Reviewer #1 (highlighted by blue) Comments #1) Why do existing solutions not satisfy you? What is wrong with them, specifically? What are the disadvantages of those approaches? Answer) Previous studies on the cleaning robots are for special purposes. There are limitations to applying to the outer-wall cleaning robot for the building. a. Too high pressure for the cleaning. The devices introduced in the previous study used the very high pressure over 100 bar for special purposes. It is too high pressure to be applied to a cleaning robot for the exterior wall of a building. If it applied to exterior wall cleaning, there is a risk of damaging the wall such as the window glass sealer or decorations on the surface. b. Robot orientation and control The higher the pressure is injected, the higher the reaction force is generated. in case of disturbance, it makes very difficult to control the orientation of the robot. Also The high reaction force ask high thrust force to attach the robot on the wall. In order to overcome the high-pressure injection reaction force, a larger lift-force generator (propellers & motor) must be used and resulting in a high specification. So it causes a significant control difficulties between high pressure and thrust force. c. Expensive equipment and weight Equipment that generates high pressure is quite large in size and increases in weight. Therefore, the burden on the rope is large to use the ascender device. In addition, as the weight increases, the performance of the posture control for dwarf eggs deteriorates. d. cost Most of the high-pressure devices are expensive. In addition, the size of devices are bigger than we think normally. if high pressure spray units is adopted, the robot size becomes bigger as well. it also ask the bigger propeller and more powerful motor for the thrust force, and robot frame and safety device units. For this reason, it becomes a cause of an increase in manufacturing cost in use. Also, in the case of existing commercial robots, there also critical disadvantages that are not satisfactory. The disadvantages of the existing method are also as follows. a. Dead zone generated during cleaning process. In the case of using a existing adopted device, such as an brush or squeeze, there is a limitation to the cleaning quality in the process of overcoming obstacles as shown in

the figure below. To prevent the collision between cleaning device and the building parts, optimized obstacles-overcoming-path are required. From the overcoming path, uncleaned area(dead zone) occurs as shown following figures. In the case of using squeeze [Reference], it is not only unfavorable for cleaning in the dead zone, but also impossible to clean continuously at the junction of the glass wall. b. Commercialization The robots adopted a squeeze or a brush are consisted more complex than high pressure water spraying device. This is because mechanical devices are required to be constantly attached to the wall. so many parts are need to keep the distance (a) as shown in figure. (In proposed method, the distance (b) is not important factor because small distance change dose not effect to the cleaning performance. ) In the view of commercialization, the complicated device have the difficulties of manufacturing the equipment and being required the high cost of the product price and the high maintenance fee. So not good for the commercialization. c. Prevention of damage to the building surface. Foreign objects-such as small stone-in the brush or squeeze can cause continuous damage to the building's surface. It can be a serious fault for the cleaning robot, especially on painted surfaces. Therefore, the proposed method can increase the quality of cleaning performance better in the dead zone than the existing method, and is advantageous in terms of production cost and maintenance as well as commercialization. In addition, we thought that it is a way to overcome the shortcomings of the existing devices as it can reduce damage to the cleaning surface by foreign substances. Proposed cleaning device(High pressure W. ) Existing cleaning device(Brush type) Real moving path obstacle Cleaning device a Dead zone obstacle Ideal moving path Nozzle High pressure water a Cleaning device b Nozzle Dead zone b Ideal moving path Real moving path High pressure water

Before correction To maximize the cleaning performance, optimization of the design parameters is required. Until now, several high-pressure cleaning researches have been conducted to enhance the cleaning performance. Zhang reported that the shape characteristics of the nozzle can enhance the high-pressure cleaning performance [19]. Peng analytically investigated the effect of the spraying distance of the nozzle on the cleaning performance [20] and Yang studied the relation of the spray pressure and installation of nozzle angle in the spray system [21]. A theoretical model for evaluating the optimal and critical distance for cleaning using high-pressure water was proposed by Guha [22]. Xu and Chen introduced the studies of the spray properties and the effect on nozzles at high pressures [23, 24]. Medan optimized the cleaning equipment to measure the reaction force of the spraying nozzle and to derive the main factor of high-pressure cleaning [25]. Zhong studied the surface erosion according to nozzle type [26], and Sertore examined the state of the nozzle using the injection force measured by load cell [27]. However, in our view, the above researches does not satisfy us as solutions for cleaning the surfaces of high-rise buildings, so further research is needed to enhance the cleaning performance. After correction Several studies on high-pressure cleaning have been conducted to enhance the cleaning performance through spray optimization. Zhang reported that the shape characteristics of the nozzle can enhance the performance of high-pressure cleaning [19]. Peng analytically investigated the effect of the spray distance of the nozzle on the cleaning performance [20], and Yang studied the effect of the installation of nozzle angle on the spray pressure in the spray system [21]. A theoretical model for evaluating the optimal and critical distances for cleaning using high-pressure water was proposed by Guha [22]. Xu and Chen pioneered the studies on spray properties and their effect on nozzles at high pressures [23, 24]. Medan optimized a cleaning equipment to measure the reaction force of the spray nozzle and to determine the main factor that affects high-pressure cleaning [25]. Zhong studied surface erosion according to nozzle type [26], and Sertore examined the state of a nozzle based on the injection force measured by a load cell [27]. However, these research studies employed very high pressure for special purposes; the findings are therefore not applicable in cleaning the outer walls of buildings. This is because high pressure has certain limitations in terms of surface damage, robot orientation control, and commercialization.

Reviewer #1 (highlighted by blue) Comments #2) How about qualitative analysis of performance? Answer) Thank you for the good question. If a robot can do the same as a human cleaning, it is best to squeeze it with squeeze with the detergent. Also, in an interview with an long experienced building exterior cleaning worker, the best way to get cleaning quality is to use squeeze and a special detergent when cleaning the exterior wall of a building, even it can easily remove scale or oil based dust created by water. However, if a machine or robot do a cleaning with squeeze, there are many mechanical limitations to cleaning. That is the why most cleaning robots use brushes. So it would be reasonable to compare and analyze the cleaning quality between brush and highpressure spray. In order to compare the quality and performance of proposed cleaning device, Additional test was conducted. In the test, similar contaminant conditions was used (reference [29]) and evaluated as follows. As can be seen from the test results, the results between a brush and with high pressure spring water were almost comparable with the visual inspection. Cleaning test with brush( reference [29] ) Cleaning test with high pressure water(Proposed) 3 bar(300 mm) 6 bar(300 mm) 8 bar(300 mm) In a cleaning performance verification test for the frame that dust left for a long time, it was confirmed that the angle did not significantly affect the cleaning performance. Also, we could identify that spraying angle prevents re-contamination as shown the following figures. Spraying angle: 45º, 8 bar, 200 mm Spraying angle: 70º, 8 bar, 200 mm Uncleaned area (Dust) Cleaned area High pressure Spring water

Reviewer #1 (highlighted by blue) Comments #3) What are the effects of wheel damage on the buildings that your device may cause? Answer) In the early stage of the robot design, the issue of building surface damage by the wheel was reviewed. The main mechanism of the damage was a scratch by the slip between the wheel and surface. So in order to prevent the slip, the robot currently being developed adopted ball bearings to get the better rotation performance and to reduce the rotating resistance of the wheel. Also the soft rubber material and airinjected type wheel was applied to get good contact characteristics than a solid wheel. Sky Pro[11 -12] IPC Eagle[9 -10] Air-injected type of wheel [*] Existing commercial cleaning robots such as IPC Eager, Skypro use the wheel of soft rubber material on the wheel set for the same issue. Another advantage of the airinjected wheel is that it can damp the impact force caused by the robot during cleaning. But it may not be enough to compensate the impact force with wheel only while the robot falls off from the surface and then land on the surface. So to reduce this, the robot under development has adopted a suspension system. Through the suspension device, It is designed to minimize the impact force effect on the contact point. [*] http: //casternara. com/index. htm Before correction For travelling on the wall, triangular wheel sets like star wheel of MSRox[34] (each wheel diameter is 110 mm) are applied, and can overcome the obstacles lower than 100 mm in height, such as window frames. Here, unlike conventional robots, the robot sprays high pressure water continuously during overcoming obstacles. It has another important advantage of minimizing cleaning dead zone compared to existing cleaning. In case of higher than 100 mm of obstacles, a separate mechanism is needed. So we have been researching another rotating-spoke-wheel(RSW) module mechanism, which can overcome 300 mm obstacle-height, and the RSW will be applied and verified the mechanism through the field test in the near future After correction Triangular wheel sets such as star wheel of MSRox [34] (each wheel with a diameter of 150 mm) are applied for traversing on walls. These can overcome obstacles with heights less than 100 mm, such as window frames. To prevent damage to the building’s surface by slippage, wheels made of a soft material were applied. In addition, a suspension device and air-injected wheels were adopted to minimize the transmission of impact force and to protect the surface. Unlike conventional robots, this robot sprays high-pressure water continuously while overcoming obstacles. This is an important advantage for reducing the amount of uncleaned zone compared with that for existing cleaning robots.

Reviewer #1 (highlighted by blue) Comments #4) "In case of higher than 100 mm of obstacles, a separate mechanism is needed. " -> Isn't this the problem you are solving over existing approaches, i. e. , not being hindered by obstacles without using separate mechanisms? Answer) This paper is the study on the optimal conditions of the high-pressure spraying for outer-wall cleaning. As reviewer mentioned, the high height obstacle overcoming mechanism is not related with the subject directly. Therefore, the sentence that reviewer pointed out was revised as follows. Before correction For travelling on the wall, triangular wheel sets like star wheel of MSRox[34] (each wheel diameter is 110 mm) are applied, and can overcome the obstacles lower than 100 mm in height, such as window frames. Here, unlike conventional robots, the robot sprays high pressure water continuously during overcoming obstacles. It has another important advantage of minimizing cleaning dead zone compared to existing cleaning. In case of higher than 100 mm of obstacles, a separate mechanism is needed. So we have been researching another rotating-spoke-wheel(RSW) module mechanism, which can overcome 300 mm obstacle-height, and the RSW will be applied and verified the mechanism through the field test in the near future. . After correction For travelling on the wall, triangular wheel sets like star wheel of MSRox[34] (each wheel diameter is 150 mm) are applied, and can overcome the obstacles lower than 100 mm in height, such as window frames. In order to prevent damage to the building surface by slip, the wheel made of a soft material were applied. In addition, to minimizes the impact force transmission and protect the surface, the suspension device and air-injected wheel are adopted. Unlike conventional robots, the robot sprays out the high pressure water continuously during overcoming obstacles. It is important advantage of minimizing uncleaning zone compared to existing cleaning robot.

Reviewer #1 (highlighted by blue) Comments #5) - "In other words, the further away the distance, the poorer the cleaning performance. " -> Isn't this expected? Answer) As the reviewer pointed out, the further away the distance, the lower the cleaning performance can be easily predicted like a common sense. Also the meaning of the sentence is similar with the previous sentence. the sentence is omitted as follows. Before correction The result of sensitivity analysis of the selected design variables is shown in Figure 8, which is calculated via Equation (2). The optimal combination of the design variables was driven as follows: an 8 -bar nozzle inlet pressure, a 0. 2 m spraying distance, a 40º spraying angle, and a 3 m/min transfer speed of the specimen. The most sensitive parameter was the nozzle inlet pressure, and the least sensitive parameter was the spraying angle. As the pressure is increased, the cleaning performance increased. Additionally, the spraying distance exhibited an inverse relationship with the cleaning performance. In other words, the further away the distance, the poorer the cleaning performance. After correction The result of the sensitivity analysis of the selected design variables is shown in Figure 8. This is calculated via Equation (2). The optimal combination of the design variables was determined as follows: nozzle inlet pressure = 8 bar, spray distance = 0. 2 m, spray angle = 40°, and transfer speed of the specimen = 3 m/min. The highest and lowest sensitivities were to the nozzle inlet pressure and spray angle, respectively. The cleaning performance increased as the pressure increased. In addition, the spray distance exhibited an inverse relationship with the cleaning performance.

Reviewer #1 (highlighted by blue) Comments #6) Why weren't obstacles used in the evaluation? Answer) As reviewer pointed out, evaluating cleaning ability while overcoming obstacles is a very meaningful. It should be proceed to show the merit of proposed method. However, the test to check the cleaning performance while overcoming obstacles must be done with control of other devices. In particular, since it is moved by a rope, the control precision and position accuracy of the ascender must be secured firstly. As reviewers know well, because small mistakes can lead to unpredictable results, it is essential to ensure the reliability of the control. In the near future, the optimized ascenders with differential gear set for the precision position control is secured, it will be evaluated in the lab-test. Before correction We confirmed the optimized conditions through a confirmation test: a nozzle inlet pressure of 8 bar, a spraying distance of 0. 2 m, and a nozzle injection angle of 40° with 3 m/min descending speed. The reaction force and impact force generated by the high pressure were 2. 6 N and 2. 7 N respectively, it is not large enough to affect the propeller thrust or damage the building surface. Based on the result of this research, a field test is carried out using a more complete cleaning device mounted on the cleaning robot. And then, further research on the optimization of the cleaning system and algorism will be conducted for abnormal conditions. For example, a study on the cleaning algorithm when overcoming obstacles and the cleaning algorithm when controlling posture. Before correction We verified the optimized conditions through a test under the following conditions: nozzle inlet pressure = 8 bar, spray distance = 0. 2 m, nozzle injection angle = 40°, and descending speed = 3 m/min. The reaction force and impact force generated by the pressure were 2. 6 N and 2. 7 N, respectively. These are not adequate to affect the propeller thrust force or damage the building surface. In the near future, based on the results of the proposed method, lab tests are expected to be carried out with the upgraded ascender control algorithm for overcoming obstacles. Through this test, the cleaning quality of the high-pressure-water spray while overcoming obstacles would be verified, and the uncleaned zone generated by obstacles would be assessed. Thus, a study will be conducted to reduce the uncleaned zone and thereby achieve good cleaning quality. Subsequently, a field test will be conducted to verify the cleaning performance of the proposed cleaning method.

Reviewer #1 (highlighted by blue) Comments #7) Why does slow speed impede performance? Isn't this counterintuitive? Answer) As the reviewer mentioned, the slower the speed, the better the cleaning performance. This is because, at a slow speed, a lot of high pressure water per unit time transfers a lot of impact energy to remove the dirt. So the quality of cleaning is very good at the low speed. As pointed out by the reviewers, the sentences “And the slower moving speed of the robot, the poorer the cleaning performance” are practically wrong. It is conflicting with the context before and after the sentence. It seems that the meaning was distorted due to the omission of the condition of the experimental content in the sentence. In order to clarify the meaning of the sentence, it was modified as follows. - In the text. The SNR was calculated using the experimental data via Equation (2). As indicated by the test results in Table 4, the cleaning performance with a high pressure was better than that with a low pressure, and the cleaning performance at a low descending speed was better than that at a high descending speed. But as shown in the experimental results of Figure 9 g) -l), the robot's descent speed was not more sensitive than nozzle inlet pressure, spraying distance, and spraying angle. Before correction In this study, a method spraying high pressure water for cleaning the outer wall of a building was developed, and the cleaning performance was optimized. The proposed cleaning device is installed on a robot equipped with a winch. Cleaning is performed by spraying high-pressure water using nozzles, a high-pressure pump, and a guide wall to prevent recontamination. The control parameters were optimized utilizing the Taguchi optimization method for different descending speed. The design variables were derived from experiences to evaluate the cleaning efficiency through image handling processing. We carried out the sensitivity analysis to identify the most sensitive design variables for the cleaning performance. The effect of the variables on the cleaning performance decreased in the following order: the nozzle inlet pressure, spraying distance, and injection angle, as shown as Figure 8. And the slower moving speed of the robot, the poorer the cleaning performance. Before correction In this study, a method of spraying high-pressure water for cleaning the outer walls of buildings was developed, and the cleaning performance was optimized. The proposed cleaning device was installed on a robot equipped with a winch. Cleaning was performed by spraying high-pressure water using nozzles, a high-pressure pump, and a guide wall to prevent recontamination. The control parameters were optimized by utilizing the Taguchi optimization method for different descending speeds. The design variables were identified based on experience to evaluate the cleaning efficiency through image processing. We carried out a sensitivity analysis to identify the design variables to which the cleaning performance is most sensitive. The following are the variables arranged in decreasing order of influence on the cleaning performance: nozzle inlet pressure, spray distance, and injection angle. The cleaning performance was not good for a slow speed of motion under the condition of a small spray angle or low spray pressure.

Reviewer #1 (highlighted by blue) Comments #8. Minor detail) The paper is not really about a study. Instead, an approach is proposed; the study helps evaluate the proposal. I suggest changing the title to something like: "An Automated Technique for High-pressure Water-based Window Cleaning and Accompanying Parametric Study" (the latter half about the study may be dropped). Answer) Thank you for the good suggestion and the correction for the title. As reviewer suggested, the title was revised as follows. Before correction Parametric study on the automated high-pressure water-based window cleaning device Before correction Automated technique for high-pressure water-based window cleaning and accompanying parametric study

Reviewer #1 (highlighted by blue) Comments #9. Minor detail) The motivation is that many high-rise buildings need cleaning. However, with COVID 19, many of these buildings are empty. COVID-19 may have a significant impact on commercial real estate. How does this affect the impact of the proposed approach? Are there other contexts where the device can be used in light of COVID-19? Answer) This comment seems to be on an economic view of the cleaning robot industry. As reviewer mentioned, purchasing or managing a cleaning robot is the economic burden when the real estate economy is the depression like nowadays covid 19 era. So it is possible to increase the cleaning cycle or even not clean the exterior walls to reduce maintenance costs. However, it is unrealistic story not to clean it for too long time. Because it asks the higher the cost as much as the period not to clean in the next cleaning. The reason that people feel the economic burden of cleaning in building maintenance is a human do the cleaning In a high-risk environment. However, if cleaning machine or robot is used on behalf of humans, the cleaning cost become lower than the previous. Moreover, if the robot is simpler and more efficient than the existing cleaning robot, It is even more deserving to operate a robot. Therefore, it can show advantages of the robot operation in a difficult real estate economy.

Reviewer #1 (highlighted by blue) Comments #10. Minor detail) - The abstract is too long—too many details. - The conclusion presents new information not found in the introduction (recontamination). The opening and conclusion should mirror each other. If this information is essential, it should also be included in the introduction. Otherwise, it should only be in the discussion section of the experimentation. Answer) Thanks for the wonderful advise and the correction. As the reviews comments, Many details in the abstract are reduced. Also the contents of re-contamination is deleted in the conclusion and abstract in order to have the unity of the text. Therefore, the abstract was modified as follows. Before correction As many high-rise buildings have been constructed recently, the maintenance of buildings has become an important issue. In the case of cleaning the outer wall of a high-rise building, a worker rides on a gondola or hangs on a rope to clean façades. This work is done in a very dangerous environment for a long time, and many accidents occur every year. Various robots are being researched and developed to reduce these incidents and to relieve workers from dangerous tasks. Herein, we propose a method of spraying high-pressure water by using a pump and nozzle, without using a brush, squeeze, or manipulator. The cleaning performance parameters, such as the water pressure, spray angle, and spraying distance, were optimized using the Taguchi method. Cleaning experiments were performed by artificially polluting window specimens. The 6 performance of the proposed method was evaluated using the image-evaluation method. From the sensitive results of image dates, the optimum condition was confirmed that a nozzle inlet pressure is 8 bar, a spraying distance is 0. 2 m, and a nozzle injection angle is 40°. The reaction force on the nozzle tip and impact force on the specimens were respectively 2. 6 N and 2. 7 N, which is not large enough to affect the propeller thrust or damages to the building surface. Based on the output of this research, a field test is carried out in the near future. After correction The maintenance of buildings has become an important issue with the construction of many high-rise buildings in recent years. However, the cleaning of the outer walls of buildings is performed in highly hazardous environments over long periods, and many accidents occur each year. Various robots are being studied and developed to reduce these incidents and to relieve workers from hazardous tasks. Herein, we propose a method of spraying high-pressure water using a pump and nozzle, which differs from conventional methods. The cleaning performance parameters, such as water pressure, spray angle, and spray distance, were optimized using the Taguchi method. Cleaning experiments were performed on window specimens that were contaminated artificially. The cleaning performance of the proposed method was evaluated using the image-evaluation method. The optimum condition was determined based on the results of a sensitive analysis performed on the image data. In addition, the reaction force due to high pressure and impact force on the specimens were investigated. These forces were not sufficient to affect the propeller thrust or cause damage to the building’s surface. We expect to perform field tests in the near future based on the output of this research.

Reviewer #1 (highlighted by blue) Comments #11. Minor detail) No future work section. Highlighting future work is crucial since you are planning future studies. Answer) Thanks for the wonderful advise. Before correction 6. Conclusion In this study, a method spraying high pressure water for cleaning the outer wall of a building was developed, and the cleaning performance was optimized. The proposed cleaning device is installed on a robot equipped with a winch. Cleaning is performed by spraying high-pressure water using nozzles, a high-pressure pump, and a guide wall to prevent recontamination. The control parameters were optimized utilizing the Taguchi optimization method for different descending speed. The design variables were derived from experiences to evaluate the cleaning efficiency through image handling processing. We carried out the sensitivity analysis to identify the most sensitive design variables for the cleaning performance. The effect of the variables on the cleaning performance decreased in the following order: the nozzle inlet pressure, spraying distance, and injection angle, as shown as Figure 8. And the slower moving speed of the robot, the poorer the cleaning performance. We confirmed the optimized conditions through a confirmation test: a nozzle inlet pressure of 8 bar, a spraying distance of 0. 2 m, and a nozzle injection angle of 40° with 3 m/min descending speed. The reaction force and impact force generated by the high pressure were 2. 6 N and 2. 7 N respectively, it is not large enough to affect the propeller thrust or damage the building surface. Based on the result of this research, a field test is carried out using a more complete cleaning device mounted on the cleaning robot. And then, further research on the optimization of the cleaning system and algorism will be conducted for abnormal conditions. For example, a study on the cleaning algorithm when overcoming obstacles and the cleaning algorithm when controlling posture. After correction 6. Conclusion and future works In this study, a method of spraying high-pressure water for cleaning the outer walls of buildings was developed, and the cleaning performance was optimized. The proposed cleaning device was installed on a robot equipped with a winch. Cleaning was performed by spraying high-pressure water using nozzles, a high-pressure pump, and a guide wall to prevent recontamination. The control parameters were optimized by utilizing the Taguchi optimization method for different descending speeds. The design variables were identified based on experience to evaluate the cleaning efficiency through image processing. We carried out a sensitivity analysis to identify the design variables to which the cleaning performance is most sensitive. The following are the variables arranged in decreasing order of influence on the cleaning performance: nozzle inlet pressure, spray distance, and injection angle. The cleaning performance was not good for a slow speed of motion under the condition of a small spray angle or low spray pressure. We verified the optimized conditions through a test under the following conditions: nozzle inlet pressure = 8 bar, spray distance = 0. 2 m, nozzle injection angle = 40°, and descending speed = 3 m/min. The reaction force and impact force generated by the pressure were 2. 6 N and 2. 7 N, respectively. These are not adequate to affect the propeller thrust force or damage the building surface. (Continued Next page)

After correction (continued) In the near future, based on the results of the proposed method, lab tests are expected to be carried out with the upgraded ascender control algorithm for overcoming obstacles. Through this test, the cleaning quality of the high-pressure-water spray while overcoming obstacles would be verified, and the uncleaned zone generated by obstacles would be assessed. Thus, a study will be conducted to reduce the uncleaned zone and thereby achieve good cleaning quality. Subsequently, a field test will be conducted to verify the cleaning performance of the proposed cleaning method.

Reviewer #2 (highlighted by Green) Comment #1) The research problem identified and discussed in section 1 is understood, the authors are encouraged to elaborate more on the problem formulation, with clear justification of the need for their research. Answer) Thanks for the advise for clear justification for this study. The need of the proposed study was emphasized by the limitations of existing cleaning robots and adding them to the text. Especially, the text has been revised as follows that the existing cleaning methods have a limitation in which uncleaned areas are created in the process of overcoming obstacles. Before correction Many exterior wall-cleaning robots have been researched [1 -7] and developed to reduce the economic cost of cleaning and to relieve workers from dangerous tasks. For example, CSIC’s Tito [8], IPC Eagle’s High. Rise [9, 10], and Sky. Pro [11, 12] are commercially available with some product lineups, as shown in Figure 1. These robots are transported up and down the building by cables through a winch. Another famous robot is Gecko [13 -15], which was developed using a powerful vacuum suction pad [16 -18]. The robots use the typical equipment to clean the exterior walls of the building, such as a water nozzle, squeegee, and brush [8 -12]. Although this is a very effective method for cleaning, continuous cleaning is impossible because of complicated building shapes and obstacles such as façades. A mechanism for overcoming these obstacles is required. This causes the robot to become complicated, increasing the price of the equipment and the maintenance cost, which is disadvantageous for commercialization. After correction Many exterior wall-cleaning robots have been studied [1 -7] and developed to reduce the cost of cleaning and to relieve workers from hazardous tasks. For example, CSIC’s Tito [8], IPC Eagle’s High. Rise [9, 10], and Sky. Pro [11, 12] are commercially available with certain product lineups, as shown in Figure 1. These robots are transported up and down buildings by using cables and winches. These commercial robots use typical equipment to clean the exterior walls of buildings, such as water nozzles, squeegees, and brushes [8 -12]. Although these devices are highly effective for cleaning, their use results in uncleaned zones. In the process of traversing optimized moving paths, which prevent collision of the devices with obstacles, uncleaned zones are generated near obstacles such as façades. Another popular robot is Gecko [13 -15], which was developed using a highly effective vacuum suction pad [16 -18]. However, the applicability of these robots are limited owing to the shapes of the buildings and the decorative fittings on them. The motion speeds are extremely low when vacuum pads are used; therefore, long cleaning times result. In addition, there are limitations on the payload required for the cleaning device. .

Reviewer #2 (highlighted by Green) Comment #2) The authors mention some relevant studies, but do not go in length in delineating how past studies and precedents have failed to achieve the desired objective regarding exterior wall cleaning performance (more in terms of metrics and data, rather than anecdotal information). This should help ground more the research and justify the discussed methods and procedures later on. Answer) Thank you for the good opinion. This comment is similar to Comment #1 of Reviewer #1. Therefore, it was modified as follows. The cited studies are for special purposes such as a cleaning the inside of drilling pipes. So these studies are about high pressures of hundreds of bars or more. It is too high to apply to the exterior wall cleaning. Because the high pressure effects the stability of robot's orientation by the reaction force and cases the damage of building's surface by the impact force generated by spraying the high pressure water. As mentioned by the reviewer, it was not clear and not enough to justify the need of this research in the text. In order to clarify the necessity of the study, it was modified as follows. Before correction To maximize the cleaning performance, optimization of the design parameters is required. Until now, several high-pressure cleaning researches have been conducted to enhance the cleaning performance. Zhang reported that the shape characteristics of the nozzle can enhance the high-pressure cleaning performance [19]. Peng analytically investigated the effect of the spraying distance of the nozzle on the cleaning performance [20] and Yang studied the relation of the spray pressure and installation of nozzle angle in the spray system [21]. ]. A theoretical model for evaluating the optimal and critical distance for cleaning using high-pressure water was proposed by Guha [22]. Xu and Chen introduced the studies of the spray properties and the effect on nozzles at high pressures [23, 24]. Medan optimized the cleaning equipment to measure the reaction force of the spraying nozzle and to derive the main factor of highpressure cleaning [25]. Zhong studied the surface erosion according to nozzle type [26], and Sertore examined the state of the nozzle using the injection force measured by load cell [27]. However, in our view, the above researches does not satisfy us as solutions for cleaning the surfaces of high-rise buildings, so further research is needed to enhance the cleaning performance. After correction To maximize the cleaning performance, optimization of the design parameters is required. Until now, several high-pressure cleaning researches have been conducted to enhance the cleaning performance. Zhang reported that the shape characteristics of the nozzle can enhance the high-pressure cleaning performance [19]. Peng analytically investigated the effect of the spraying distance of the nozzle on the cleaning performance [20] and Yang studied the relation of the spray pressure and installation of nozzle angle in the spray system [21]. A theoretical model for evaluating the optimal and critical distance for cleaning using high-pressure water was proposed by Guha [22]. Xu and Chen introduced the studies of the spray properties and the effect on nozzles at high pressures [23, 24]. Medan optimized the cleaning equipment to measure the reaction force of the spraying nozzle and to derive the main factor of highpressure cleaning [25]. Zhong studied the surface erosion according to nozzle type [26], and Sertore examined the state of the nozzle using the injection force measured by load cell [27]. However, these previous researches used very high pressure for special purposes. So it is difficult to use for the cleaning for building outer surface, because the high pressure has some limitations in views of the surface damage, robot orientation control, and commercialization.

Reviewer #2 (highlighted by Green) Comment #3) There also some strong assumptions in the third paragraph in page 3 that require some clarification. If some of the cleaning devices and robots are to be adopted, there needs to be some more explanation regarding specific relations to the proposed methods described later on and how they fit together. This is missing. Answer) Thanks for the advise. The proposed cleaning device in this study was applied to a robot using a rope(the ascender). The robot is insensitive to the reaction force from the spraying pressure and the force balance according to each nozzle pressure. So the operation stability is better than that of the robot using the suction plate. In addition, there are no major issues to be solved, such as a suction problem of water and air simultaneously, a payload for devices and a damage to the adsorption plate caused by foreign matter and wear. In particular, it is advantageous on a space for adopting high pressure spraying in the total robot layout. The above contents were added in the text as follows. Before correction The robot adapted the high-pressure cleaning module will be installed is shown in Figure 2. The robot moves on the building surface using two ropes which is secured on the top of the building. To secured the ropes, separate devices are used on both side of the top building edges. This has a great advantage of simple installation and operation of the robot even in buildings without gondola or separate winch devices. Basically by two winches on the robot, it moves on the building surface. The robot gets the force for attaching the wall using the propeller thrust. After correction A robot installed in a high-pressure cleaning module is shown in Figure 2. The robot can move over the building’s surface using two ropes that are secured at the top of the building. Additional devices are installed on both edges of the building’s top to secure the ropes’ ends. This system involves simple installation and enables the operation of robots for buildings without the use of gondolas or separate winch devices. The robot uses two winches that are embedded into it. The force for attaching on to the wall is provided by the propeller. This type of robot system is relatively insensitive to the reaction force against the spray pressure and the force balance of the nozzles’ pressure. Therefore, the operational stability is higher than that of a robot that uses a suction plate. In addition, severe issues such as the occurrence of simultaneous suction of water and air, the required payload for the devices, and damages to the adsorption plate caused by foreign matter and wear do not arise. In total layout of the robot, it is advantageous on a space efficiency because of the simple structure of the cleaning unit.

Reviewer #2 (highlighted by Green) Comment #4) Section 3 lays out well some of the specifications for the proposed device, but requires some more explanation, especially in the second paragraph discussing the selection criteria. There appear to be some gaps in the description of the selection criteria. This requires further elaboration. Answer) The description meaning of the nozzle selection criteria was not clear in the text, as reviewer mentioned. In particular, in terms of impact efficiency and uniformity, the reason for selecting a flat type nozzle was unclear. So the sentences were modified to clarify the meanings as follows. Before correction The nozzle tip selection criteria were based on 2 points: the impact efficiency, which is cleaning performance factor, and the uniformity of flow rate, which determines the uniformity of cleaning performance. From the data sheets of the impact efficiency at 50º on the spray angle, the flat type was 10%, while the full cone and the hollow cone were less than 1% and 1%, respectively [35, 36]. The spray uniformity of flow rate according to the nozzle type was confirmed under the specified pressure. The flat type of nozzle has better spray uniformity with a fan shape and is the most widely used type of nozzle for cleaning. Better impact efficiency than hollow cone or full cone type of nozzle. Finally, HM_V type (Hanmi Nozzle co. ltd [35]) of the flat nozzle was selected, as shown in Figure 3. For reliability, we selected a single unit nozzle made of stainless-steel. After correction The nozzle tip selection criteria were based on 2 points: the impact efficiency of spraying water, which is the ability of removing the contaminant, and the uniformity of water spraying shape at the nozzle tip, which determines the uniform cleaning performance. From the data sheets provided by the manufacturer, the impact efficiency of the flat type nozzle at 50º was 10%. While the full cone and the hollow cone were less than 1% and 1%, respectively [35, 36]. The spraying shape uniformity of according to the nozzle type was able to confirm at the designated pressure. The flat type of nozzle has better spray uniformity with a fan shape and is the most widely used type of nozzle for cleaning. Better impact efficiency than hollow cone or full cone type of nozzle. Finally, HM_V type (Hanmi Nozzle co. ltd [35]) of the flat nozzle was selected, as shown in Figure 3. For the reliability, a single unit nozzle made of stainless-steel selected.

Reviewer #2 (highlighted by Green) Comment #5) Most of the figures, especially 4, 5, 9, and 10 are unreadable and difficult to read. These are crucial illustrations that should be much more developed and more clear, perhaps enlarged in size, to demonstrate the suggested differences in cleaning performance, and so they require much more clear images and image resolution. Answer) As the reviewer pointed out, the clarity of each picture was improved by changing the format of all pictures. And the figure size was increased so that the results of the experiment could be easily and clearly confirmed. In addition, in the case of ambiguous pictures, it has been modified as follows. Before correction Figure 4. (a) Control factors of cleaning performance. A: Nozzle inlet pressure, B: Spraying distance, C: Spraying angle. (b) User condition, Cleaning speed of the robot. After correction c) b) Nozzle Applied pressure water (A) High pressure water Spraying distance (B) a) d) Cleaning velocity (D) Spraying angle (C) Surface Figure 4. Control parameters (a–c) and a user condition (d). a) Nozzle inlet pressure, b) Spray distance, c) Spray angle. d) Cleaning speed of the robot. [29, 37].

Reviewer #2 (highlighted by Green) Comment #5) Most of the figures, especially 4, 5, 9, and 10 are unreadable and difficult to read. These are crucial illustrations that should be much more developed and more clear, perhaps enlarged in size, to demonstrate the suggested differences in cleaning performance, and so they require much more clear images and image resolution. Continue answering of comment #5) As reviewer mentioned, the images were changed to much more clear images and additional indexes were added. Before correction Figure 5. Measurement setup for the cleaning-performance evaluation: a) setup for taking photographs of the acryl panel; b) panel before cleaning; c) panel after cleaning; d) converted visual data from the photographs [29, 37]. After correction a) b) c) Tripod Camera LED Lamp Figure 5. Device setup for the image data measurement and conversion to image data from the test results. a) Device setup to capture photographs for the image data, b) A test result photograph after cleaning, c) Image data converted from the photographs [29, 37].

Reviewer #2 (highlighted by Green) Comment #5) Most of the figures, especially 4, 5, 9, and 10 are unreadable and difficult to read. These are crucial illustrations that should be much more developed and more clear, perhaps enlarged in size, to demonstrate the suggested differences in cleaning performance, and so they require much more clear images and image resolution. Continue answering of comment #5) To help understanding of the experimental process, the images were changed to clearer images. Before correction Figure 7. Experimental test bench setup used for measuring the cleaning performance. a) test-scene of high pressure water cleaning, b) Unit A: nozzle spraying unit, c) Unit B: Cleaning performance evaluation unit and transfer devices. Specification of significant part are followings. Pump (8905 -902 -290, SHURflo), Pressure sensor: GP-M 025, KENYCE), Load cell: BCA-5(CAS), Nozzle: Flat type, HM_V 5 (Hanmi Nozzle. co. Ltd) [35]. After correction a) c)` b) z x Y Figure 7. Experimental test bench setup for evaluating the cleaning performance. a) Test scenario with high-pressure water cleaning, b) Cleaning performance evaluation unit and transfer devices in two directions (X, Y). c) Water-sprayunit assembly. Specifications of main parts: Pump: 8905 -902 -290(SHURflo), Pressure sensor: GP-M 025, (KENYCE), Load cell: BCA-5(CAS), Nozzle: Flat type, HM_V 5 (Hanmi Nozzle. co. Ltd) [35].

Reviewer #2 (highlighted by Green) Comment #5) Most of the figures, especially 4, 5, 9, and 10 are unreadable and difficult to read. These are crucial illustrations that should be much more developed and more clear, perhaps enlarged in size, to demonstrate the suggested differences in cleaning performance, and so they require much more clear images and image resolution. Continue answering of comment #5) As shown in the figure below, Figure 8 has been revised to a larger index and clearer images. Before correction 30 SNR (d. B) 20 10 0 -10 -20 3 bar 6 bar 8 bar Nozzle inlet pressure 0. 2 m 0. 3 m 0. 4 m Spraying distance 15° 25° 40° Spraying angle Figure 8. Sensitivity analysis results of the design variables based via Equation (2) After correction 30 SNR (d. B) 20 10 0 -10 -20 3 bar 6 bar 8 bar Nozzle inlet pressure 0. 2 m 0. 3 m 0. 4 m Spraying distance 15° 25° 40° Spraying angle Figure 8. Sensitivity analysis results of the control parameters based on Equation (2)

Reviewer #2 (highlighted by Green) Comment #5) Most of the figures, especially 4, 5, 9, and 10 are unreadable and difficult to read. These are crucial illustrations that should be much more developed and more clear, perhaps enlarged in size, to demonstrate the suggested differences in cleaning performance, and so they require much more clear images and image resolution. Continue answering of comment #5) As shown in the figure below, Figure 10 has been revised to a large and clearer images. Before correction Figure 9. Result of the best and worst cleaning performance. The best performance result was at Exp. #7 on Table 4 when the user condition was 3 m/min as shown in a)–c) and d)–f) were when 6 m/min. The worst performance result was at Exp. #6 when both user condition. The descending velocity was g)–i) 3 m/min and j)–l) 6 m/min. After correction a) b) c) d) e) f) g) h) i) j) k) l) Figure 9. Test results for the highest and lowest cleaning performances. The highest performance results were those of Exp. #7 (Table 2) for the user condition of 3 m/min, as shown in a)–c). d)–f) display the results for 6 m/min. The worst performance result was for Exp. #6 under both the user conditions (3 mm/s and 6 mm/s). .

Reviewer #2 (highlighted by Green) Comment #5) Most of the figures, especially 4, 5, 9, and 10 are unreadable and difficult to read. These are crucial illustrations that should be much more developed and more clear, perhaps enlarged in size, to demonstrate the suggested differences in cleaning performance, and so they require much more clear images and image resolution. Continue answering of comment #5) As shown in the figure below, Figure 10 has been revised to a large and clearer images. Before correction Figure 10. Results of confirmation experiment for the optimized cleaning variables (nozzle inlet pressure: 8 bar; spraying distance: 0. 2 m; injection angle: 40°), as well as the image data result: 74. 8% (average). After correction Figure 10. The confirmation test results under the optimal conditions. Nozzle inlet pressure: 8 bar; spraying distance: 0. 2 m; injection angle: 40°. The image data results: 74. 8% (average).

Reviewer #2 (highlighted by Green) Comment #6) a) Some figures are misplaced, such as figure 6, which needs to belong under section 4. 4. b)Table captions need to be above the tables not below. Answer) a) As shown below, the position of the figures were rearranged in an appropriate position according to the content of the text. In addition, table captions were also repositioned to the top of the table from the bottom. 4. 4 Test bench configuration and experiment setup The test bench was designed with two parts as shown in Figure 6. A water spraying unit consist of a load cell, pressure sensor and a nozzle device assembly. And a cleaning-performance evaluation unit has window frame with a load cell, transfer devices with 2 axis. In the water spraying unit, the nozzle feed pressure was the control factor. The pressure was changed by modifying the RPM of the pump and a pressure sensor monitored nozzle inlet pressure. The nozzles used in the experiment had the same orifice shape, but their spraying angles were different respectively [35]. . . . a) b) Pressure sensor Load cell Cleaning-performance evaluation unit Window frame (Acryl specimens) Manifold Nozzle tip Load cell Transfer device (2 axis) Nozzle spraying unit Pump Figure 6. Test bench consist of 2 parts: a cleaning performance evaluation unit for investigating spraying characteristics and a nozzle spraying unit for implementing movement of the robot cleaning. Each unit have a load cell to measure the reaction force and impact force. (a) Overall of test bench and (b) the water spraying unit.

Reviewer #2 (highlighted by Green) Comment #6) Some figures are misplaced, such as figure 6, which needs to belong under section 4. 4. Table captions need to be above the tables not below. Continue answering of comment #6) b) The table caption was replaced as follows. Before correction Control Parameters Level 1 Level 2 Level 3 A Spraying angle (°) 15 25 40 B Spraying distance (mm) 200 300 C Nozzle inlet pressure (bar) 3 6 Table 1. Control factor for optimization of the cleaning device User condition Level 1 Level 2 Cleaning Speed (m/min) 3 6 Table 2. User condition of cleaning robot cleaning speed Pressure (bar) 3 6 9 Flow rate(L/min) 2 2. 8 3. 4 Table 3. Relationship between the flow rate and pressure [35] After correction Table 1. Control factor for optimization of the cleaning device Parameter Level 1 Level 2 Level 3 8 Nozzle inlet pressure (bar) A 3 6 Control Parameters Spraying distance (mm) B 200 300 Spraying angle (°) C 15 25 User condition Cleaning speed(m/s) D 3 6 -

Reviewer #2 (highlighted by Green) Comment #7) The first two paragraphs in page 5 are relatively confusing in relation to the tables and figure that follow. a) Please describe very clearly the content of each table (in a more elaborate discussion) so that the flow of information reads well. b) There is no adequate discussion regarding the mentioned levels (1, 2 and 3) and how they relate each of the parameters. Answer) a) Because the parameter order between Figure 4 and Table 1 was different, it was difficult to understand the control parameters. It seems that it is a mistake in thesis writing process. The parameter have been reorganized for easy-reading flow as follows, and then the order of parameters is consistent with Figure 8. Before correction FIGURE 4. (a) Control factors of cleaning performance. A: Nozzle inlet pressure, B: Spraying distance, C: Spraying angle. (b) User condition, Cleaning speed of the robot. Table 1. Control factor for optimization of the cleaning device Control Parameters Level 1 Level 2 Level 3 40 A Spraying angle (°) 15 25 B Spraying distance (mm) 200 300 C Nozzle inlet pressure (bar) 3 6 After correction a) c) Nozzle High pressur e water Nozzle inlet pressure (A) = 3/6/8 bar Spraying distance (B) b) d) Cleaning speed(D) = 3, 6 m/min Spraying angle (C) Surface Figure 4. Control factors(a-c) and a user condition(d). a) Nozzle inlet pressure (A), b) Spraying distance (B), c) Spraying angle (C). d) Cleaning speed of the robot Table 1. Control factor for optimization of the cleaning device Control Parameters User condition Level 1 Level 2 Level 3 Nozzle inlet pressure (bar) Parameter A 3 6 8 Spraying distance (mm) B 200 300 Spraying angle (°) C 15 25 Cleaning speed(m/s) D 3 6 -

Reviewer #2 (highlighted by Green) Comment #7) The first two paragraphs in page 5 are relatively confusing in relation to the tables and figure that follow. a) Please describe very clearly the content of each table (in a more elaborate discussion) so that the flow of information reads well. b) There is no adequate discussion regarding the mentioned levels (1, 2 and 3) and how they relate each of the parameters. Continue answering of comment #5) As pointed out by the reviewer, The criteria of the level for the control parameter are summarized as follows. The control factors were divided into 3 levels. In the case of nozzle inlet pressure, since the maximum tap water pressure is generally about 2. 7 bar, the minimum spraying pressure was 3 bar. Max nozzle inlet pressure was determined as 80% of the maximum discharge pressure (10 bar) of the selected pump considering the safety margin. As for the spraying angle, based on the large impact force and most frequently used types at the cleaning site, three angle of the flat nozzles were selected. In the view of the spray angle, at more than 40 degrees, the impact force is considerably reduced, and satisfactory cleaning performance cannot be expected. As can be identified from the definition, the spraying distance is closely related to the overall layout of the cleaning robot. Not only the space between the outer wall and the nozzle tip but also the cleaning area are closely related with the robot structure. In addition, the distance is a factor that determines how high obstacles to overcome. The distance was determined from 200 mm to 400 mm in consideration of the performance and the robot-layout. Before correction Table 1 presents the design variables that need to be optimized to achieve the maximum performance, as shown in Figure 4. We selected the following three sensitive variables for highpressure water cleaning: nozzle inlet pressure, nozzle spraying distance, and spraying angle. The nozzle inlet pressure is related to the flow rate as follows [21]. Here, Q, P, and n represent the flow rate, pressure, and specific gravity of the fluid, respectively. According to Equation (1), we control the pressure instead of the flow rate, as shown in Table 3. Level 3 corresponds to the maximum working pressure of the water pumper, which is the actual operational condition of the cleaning device. The spraying distance is defined as the distance between the building surface and the nozzle tip. A shorter nozzle distance yields a greater impact force on the surface for cleaning. The spray angle is another important parameter affecting on the cleaning performance. The levels of the control parameters are selected based on a cleaning-robot structure that is affordable the robot body depth. The speed is an uncontrollable parameter under certain circumstances, to be met for all operating speeds. So two level of robot descending speed (3 and 6 m/min) were selected to confirm the best and worst condition of parameter’s combination, respectively, as Table 2 and Figure 4 shows

After correction We selected the following three sensitive variables for high-pressure water cleaning: nozzle inlet pressure, nozzle spraying distance, and spraying angle. The variables need to be optimized to achieve the maximum performance were presented in Table 1. And it is expressed schematically, as shown in Figure 4. The nozzle inlet pressure is related to the flow rate as follows [21]. Here, Q, P, and n represent the flow rate, pressure, and specific gravity of the fluid, respectively. According to Equation (1), we control the pressure instead of the flow rate, as shown in Table 1. As for the spraying angle, based on the large impact force and most frequently used types at the cleaning site, three angles of flat nozzle were selected. In the view of the cleaning performance, the impact force is considerably reduced at more than 40°, and satisfactory cleaning performance cannot be expected. The spraying distance is defined as the distance between the building surface and the nozzle tip. The distance is closely related to the overall layout of the cleaning robot. Not only the space between the outer wall and the nozzle tip but also the cleaning area are closely related with the robot structure. In addition, the distance is a factor that determines how high obstacles to overcome. It was determined from 200 mm to 400 mm in consideration of the performance and the robot-layout.

Reviewer #2 (highlighted by Green) Comment #8) Table 2 and 3 are quite short and their inclusion is questioned. Perhaps another tabulation format can be included, otherwise they could be eliminated and included as text description, but with further explanation Answer) As pointed out by the reviewer, Table 2 and Table 3 do not contain much information. So Table 1, Table 2 and Table 3 were merged into one table as shown into Table 1. Before correction Control Parameters Level 1 Level 2 Level 3 40 A Spraying angle (°) 15 25 B Spraying distance (mm) 200 300 C Nozzle inlet pressure (bar) 3 6 Table 1. Control factor for optimization of the cleaning device User condition Level 1 Level 2 Cleaning Speed (m/min) 3 6 Table 2. User condition of cleaning robot cleaning speed Pressure (bar) 3 6 9 Flow rate(L/min) 2 2. 8 3. 4 Table 3. Relationship between the flow rate and pressure [35] After correction Table 1. Control factor for optimization of the cleaning device(* Flow rate corresponding to the nozzle inlet pressure) Control Parameters User condition Parameter Level 1 Level 2 Level 3 A 3 (2) 6 (2. 8) 8 (3. 4) Spraying distance: mm B 200 300 Spraying angle: ° C 15 25 Cleaning speed: m/s D 3 6 Nozzle inlet pressure: bar (L/min*) -