Washing machine and dryer use at campus Leidse Schans Mio T.L. Poortvliet December 4, 2019 Resident Committee Leidse Schans Abstract The resident committee (RC) of campus Leidse Schans collected data about washing machine and dryer use between October 27 and November 18. From this data conclusions can be drawn about peak use and careful guesses can be made towards the in- fluence of temperature and rain on the use of the laundry room. It should be noted that the infor- mation from duwo.multiposs.nl, where machine use was recorded from, does not report accurately. The laundry room is used least during 02:00-06:00 and a smaller dip in use is observed between 15:00-18:00. Rain and temperature seem to have no effect on laundry room use, but more data is needed to attenu- ate statistical fluctuations and seasonal dependence. Two solutions are proposed to increase capacity and user friendliness of the laundry facilities. Currently demand outweighs supply. 1 Introduction Campus Yours, or campus Leidse Schans, is a resi- dential project that is nearing its completion. At the time of writing the campus provides 1493 houses [1] , 1070 of which belong to student housing cooperation DUWO [2;3;4] . The DUWO buildings have the oppor- tunity to use the laundry room, this is a room on the ground floor of Omegaplantsoen that contains eight washing machines and five dryers. That comes out to one washing machine per 134 houses. Assuming every house uses one washing machine and all ma- chines are used all day, each house gets 1 hour and 15 minutes of machine time. However, this calculation is based on a flawed assumptions; it is not realistic to expect users to do their laundry in the middle of the night. Another flaw in that calculation is that a user will often have more laundry than can be done with one machine. The laundry room is often too busy in the experience of users, as is backed up by above calculation. Finding a time slot to do laundry can be a challenge. The resident committee looked at laundry room use. Data was collected of washing machine and dryer use during a period between October 27 and November 18. The feedback from residents led to the question: ‘are there enough machines for all res- idents to do their laundry?’ To answer that ques- tion properly another one needs to be answered first: ‘what are the influences on laundry room usage?’ In this document the results are shown, interpreted and based on them a conclusion is formed. In the following document the analysis of the RC is presented. Sections 2 and 3 explain and justify the way data is presented in section 4 and further. The reader can skip sections 2 and 3, they can always come back to them if they are interested. 2 Gathering data Bart Bussink set up a web scraper that gath- ers information from the online laundry portal duwo.multiposs.nl. This website displays the avail- able amount of washing machines and dryers in real time. However, there is a fluctuating discrepancy between actual machine use and the reports from duwo.multiposs.nl which is discussed in depth in sec- tion 5. This data, along with date and time of recording were logged in a database. Besides this data the temperature and probability of rain were recorded. 1 3 Methods of analysis The gathered data is analyzed using python and numeric, scientific and plotting modules. Vari- ous plots are made to show trends in the use of washing machine and dryer use. The goal is to gain insight into the behaviour of laundry room use. The data and code are publicly available at https://github.com/MioPoortvliet/Wasmachines- Leidsche-Schans. 3.1 Time dependence One of the most important factors in machine use is time. To gain insight the machine use is compared through various slices of the data set in time. 3.1.1 Use throughout the week First, the data is sorted by day and the mean num- ber of machines available at a specific day of the week at a specific time are calculated. Corrupt en- tries are removed. This plot considers machine use throughout the days of the week. The average ma- chines available on a day is shown as a function of the day, see figure 1. 3.1.2 Use throughout the day Another plot considers use throughout the day. The mean use in a time interval of 45 minutes over all days of the week is calculated. To correct for the varying times of recording and to somewhat smooth out the data an interval of 45 minutes is chosen. This is plotted in figure 2. 3.1.3 Rate of change With this processing the rate of change can also be computed, this is plotted in figure 3. This is sim- ply the differential quotient of the machines over the time. This signal is further filtered to smooth it. To do so, the moving point average of n points is taken. Note that as ∆ t becomes greater, the differen- tial quotient grows smaller because the fluctuations are on a relatively small timescale. 3.2 Length of use The cross correlation function of washing machine and dryer use is calculated and plotted directly from the original data set. Figure 4 contains the resulting plot. 3.3 Rain and temperature dependence The average available machines at a temperature and chance of rain are plotted in figure 5. The coloured surfaces represent the standard deviation in the data set. In the background the histogram of the rain respectively temperature is shown to show how much data is available. Presenting the data in this manner, it is easy to see what is going on. How- ever, information is lost. 4 Results Machine use depends on a lot of variables. Some are easy to measure such as the time of day, day of the week and weather. The analyzed data can now be examined visually and conclusions may be drawn. 4.1 Time dependence Time is an important factor in machine use. The mean use throughout the week is considered and af- ter that, the mean use throughout the day. 4.1.1 Use throughout the week First the dependence on the day of the week needs to be considered. Figure 1 depicts the average use throughout the week. The use seems to have a max- imum on Wednesday, although within one standard deviation. The weekend does not have a lot of im- pact. A possible explanation for the shape of the curve is that there is relatively little data for Wednesday. The fluctuations inherent in the duwo.multiposs.nl reports are more pronounced due to this. 4.1.2 Use throughout the day Since the dependence on day of the week is relatively small, it is justified to take the sum of all those days. Other plots were made similar to figure 2 where each 2 Figure 1: The average use of the machines on the vertical axis is shown. The horizontal axis represents a certain day of the week. The standard deviation is represented by the colored surface. The curve does not look flat, although the minimum and maximum are within one standard deviation of each other. day was plotted individually. This way of present- ing the data provides no further insight, however the shape of the curves match up to what can be seen in figure 2. This further justifies the averaging through- out the days of the week. Figure 2 shows the average use of the machines in time. Noticeably, it is very rare that all machines are available to use. The peak moment starts from 06:00 and continues until 16:00 as can be read di- rectly from figure 2. From 16:00 to 18:00 there is a slight bump in machine availability after which it decreases again. From 20:00 the availability slowly builds up again to 01:00. Noticeable is that on av- erage about two machines are always in use, even between 01:00 and 06:00. This is due to incorrect reports from duwo.multiposs.nl as is discussed in sec- tion 5. 4.1.3 Rate of change Figure 3 depicts the rate of change in machine use over time. A smaller time averaging was chosen to get a better look at the volatility of machine use. The physical interpretation of this plot is that a pos- itive rate of change corresponds to people using the washing machines. A negative rate of change cor- responds to the end of washing program, implying someone will come pick up their laundry. Thus it Figure 2: The total mean machine use during the day. Washing machines are plotted in blue and dry- ers in orange. The time of day is represented on the horizontal axis while the average available ma- chines are represented on the vertical axis. The use decreases at 21:00 until 02:00 when it reaches a min- imum. Between 06:00 and 09:00 the use picks up drastically and stays roughly constant during the day, with a small dip just before 18:00. After 21:00 the use slowly decreases back to its minimum. Figure 3: On the horizontal axis the time is repre- sented while on the vertical axis the rate of change of the machine use is represented. At the start of the series the rate of change is negative. It stays around zero from 04:00-08:00 after which it becomes positive. Between 11:00 and 19:00 oscillations can be seen. After 19:00 the rate of change abruptly decreases, peaking sharply around 23:00. It slowly trends back to zero. 3 Figure 4: The cross correlation between the washing machine and dryer use is plotted in a range between 0 to 3 hours. The horizontal axis represents time, approximately. The vertical axis is the normalized correlation function. Its maximum is at exactly 1:00 hours, which is accentuated by an orange dot. may be concluded that machine use increases from 08:00 to 11:00 after which the use peaks. A positive rate of change signifies the machine use increasing and a negative rate of change signifies a decrease of machine use. Between 11:00 and 19:00 oscillations with a period of roughly an hour are observed, this is due to the average time a cycle on the washing machine takes. 4.2 Length of washing machine use Looking at figure 2 it is noticed that the dryer use lags behind the washing machine use. The time it lags behind can be approximated using the cross correlation between the washing machine and dryer data series. This cross correlation function is cal- culated and plotted in figure 4 in the range of 0 to 36 samples, which approximately corresponds to three hours. The polling rate of the web scraper was five minutes, thus every sample translates to roughly 5 { 60 hours. This is not exact since the web scraper sometimes failed to record the data properly, how- ever it is close enough to give a good indication. Therefore the x -axis can be interpreted as a time with units of hours. At around one hour the corre- lation peaks, this can be interpreted as the average time the washing machines are used. In this correla- tion function other bumps are noticed, not shown in the plot, with a periodicity of 24 hours. This is due to the periodicity of the washing machine use from day to day, further confirming that the x -axis can be interpreted as time with units of hours. 4.3 Rain and temperature dependence To see if the laundry room use is influenced by the rain or temperature they were also recorded. Fig- ure 5a shows the average machines used dependant on temperature. As can be seen from the figure, no statistically significant temperature dependence is present. Around 0 0 C a bump is seen. Although this implies that at a colder temperature less peo- ple do laundry, this can be attributed to the lack of data in this regime. The colored surface signifying one standard deviation agrees with this hypothesis. If the events are correlated, they are not necessar- ily caused by one another; during the period when less people do their laundry, at night, it is colder. An important thing to remember is that this data is gathered in the transitional period to cold weather. By gathering data over a longer time period seasonal fluctuations will be averaged out. From figure 5b no direct correlation between rain and laundry room use can be observed. 5 Discussion First, we would like to note that the figures gath- ered from duwo.multiposs.nl are not accurate and histograms of washing machine and dryer use should be taken with a grain of salt. Some machines are unlocked and may be used without paying. These machines always show up as unavailable in duwo.multiposs.nl. For some reason, sometimes the online laundry portal reports 2 washing machines available when they are actually all in use. Further- more, we would like to point out that in the data there are times where the laundry facilities are not used as much. These periods are during the night (23:00-08:00). It is unreasonable to expect that res- idents do their laundry during this time. 5.1 Criticism of the current system As it stands, doing laundry with the machines and system available is not as user friendly as it could 4 (a) The machine use appears to be correlated with the temperature. At lower temperatures the machine use de- creases, however this is within a standard deviation. Pos- sible explanations are presented in section 4.3. (b) The machine use does not seem to depend on the rain. It fluctuates somewhat, however these fluctuations are well within a standard deviation. Figure 5: In the background, a histogram of the count of data points are shown. It is not scaled to match the axis. The lines represent the average machine use dependant on the temperature respec- tively rain. The colored surfaces once again repre- sent one standard deviation. The red line represents the washing machine use and the blue line the dryer use. be. The website duwo.multiposs.nl is slow to load, reports machine use inaccurately and the reservation system is fundamentally flawed. Reservations have no meaning since they are only valid for 15 minutes. A reservation may be made every hour on the hour. To illustrate the flaw with an example, say a reservation is made for some hour. Then, 15 minutes before the start of that hour all washing machines start a new cycle. The mean of a cycle is one hour, therefore when the reservation is valid all machines are still in use and laundry can not be done. Besides that, the machines sometimes ‘swallow your money’; after you pay in the duwo.multiposs.nl system they do not unlock. On the other hand there are machines that are always unlocked, meaning you do not have to pay to use them. Another criticism of the laundry facility is that residents have reported dirty machines and often one or more machines are not operational or do not work as intended. Our own observations line up with these reports, the machines do not seem to be maintained as well as they could be. 5.2 Unknowns in the analysis Time and weather are not the only influences on laundry room use. There are unknowns and influ- ences we cannot quantify. One such factor is that there are residents who do not use the laundry ma- chines on campus because of the problems stated above, but would like to if doing laundry with them was a less painful process. The above analysis can not take this into account, meaning that if these res- idents were to use the laundry room the presented plots would have different characteristics. 5.3 Proposed solutions The laundry facilities in their current state can not keep up with demand. An obvious solution that will solve the problem is to place another laundry room in the available room in Sigmaplantsoen. This is something that has been proposed before. DUWO has informed residents that this is not an easy oper- ation due to external constraints. The transparency from DUWO is appreciated. Another solution could be better maintenance of 5 the laundry room and a new reservation system. If all machines are operational at all times and work as intended they could keep up with demand more than they do now. In the new system users have to make a reservation to use the laundry facilities. Users will have to pay upfront to make a reservation. Then they will be guaranteed the opportunity to use the machines. However, this system relies on the fact that all machines are operational and work as intended at all times. If they do not, users will be charged money without getting to do their laundry. A criticism of this solution is that it is less efficient than the current system since there is a low duty cycle due to the reservation system. It might not increase the capacity of the laundry facilities, it does makes the process of doing laundry less frustrating. 5.4 Conclusion Currently the laundry system on campus Leidse Schans is in poor condition. To residents it feels like there are not enough machines to keep up with demand while the data tells a different story. This is due to unfaithful reports from duwo.multiposs.nl. There is small potential for more efficient use of the laundry room and demand outweighs supply by more than that. In the future, a model can be made to predict laundry use from the gathered data. This can optimize laundry room use. Corresponding Author To contact the author of this report please send an email to info@bccampusleidseschans.nl. Conflict of Interest The RC is subsidised by BRES, which is an organ of DUWO. The author is a resident of Campus Leidse Schans. Acknowledgements The author would like to thank Bart Bussink for set- ting up and running the webscraper. Bart initiated this research. References [1] “YOURS - Te huur: 423 apparte- menten in Leiden.” [Online]. Available: https://leidenisyours.nl/ [2] “Nog eens 470 studentenwoningen op cam- pus Leidse Schans.” [Online]. Available: https://www.duwo.nl/over-duwo/duwo-nieuws/ het-laatste-nieuws/nieuwsbericht/item/204/ [3] “Campus Yours (Epilsonplantsoen.” [On- line]. Available: https://www.duwo.nl/ over-duwo/onze-gebouwen/projectinformatie/ projectdetails/item/31/ [4] “Sleuteluitgifte op Leidse Schans.” [Online]. Available: https://www.duwo.nl/over-duwo/ duwo-nieuws/het-laatste-nieuws/nieuwsbericht/ item/207/ 6