Wasted Metrics
Madness in testing the reliability of resource recovery load data at UC Santa Cruz
Faced with the dramatically increasing diversion rates required for the UC Santa Cruz campus, Ground Services had a central issue with reducing landfill usage: we did not have the means to ascertain accurate, per client values for waste and recycling recovered from campus buildings. After months of research, Ground Services employed Loadman’s on-board technology to record real-time weight measurements of the materials recovered on a per client, per resource-category basis. Initially only the refuse truck was outfitted, but soon thereafter, the cardboard truck was reading service weights with the on-board scales. Conceptually, the system works with two communication systems: GPS and a cellular modem. Both systems are located within the truck to allow it to “know” where it is located while servicing dumpsters in the field. This information is then uploaded to a cloud location which is then bounced to a local database in the Grounds office. Soon, the system will be completely cloud-based.
When a dumpster is serviced, a scaling system takes measure of the container for a gross weight on the way upwards to deposit the materials in the bed of the truck. On the way down, a “tare” weight for the container is taken and subtracted from the original gross weight. By subtracting the tare from the gross, the net weight is an accurate load value for that service at the given GPS location. Prior to executing a route, “geofences” are created so that the on-board meter can automatically attribute a specific client tag to the dumpster service location at the time of pick-up. Although necessarily a complicated process, with adequate preparation and maintenance, the system is amazingly effective.
From the start, I knew insuring accurate readings was extremely important to establish both client and management confidence in the system. In order to measure the accuracy of the system, the total weight of a route as compiled in the Loadman software is compared to the known accurate weight of the resources measured on large truck scale at the resource recovery center (i.e. the dump) upon unloading. In FIGURE 3, the error rate for both our cardboard and refuse trucks is illustrated. Also shown are the upper and lower green threshold lines that indicate the +/- 5% rate we set to achieve over the course of the test period.
As you can see, by the end of the exercise, we were largely able to achieve the +/-5% accuracy range, but not without extensive review and testing in the field. The test period began August 2012 and lasted through May 2013.
Initially, we believed the error rate could be directly attributed to rain events experienced on campus. As many of the dumpsters on campus are not consistently protected from rainfall, they can become wet and weighed down in a storm. As much of the heavy rain either misses the truck bed altogether or drains out on the way to the dump, it can cause major errors in the on-board readings, typically by recording a greater accumulated weight was picked up than the one measured at the dump scale.
For refuse, error rates show graphically immediately with the rain because trash dumps are made daily. With cardboard, rain events precede error incidences because trips to the recycling center are only taken on a weekly basis. Also, since the cardboard truck is newer than the refuse truck, it has a tighter seal that does not allow rainwater to evacuate during transit. this reduced the error impact from rain events.
As a second factor tested in our error testing, we found that driver style had a significant impact on overall accuracy. With review, we found that the on-board metrics system is one that can be learned with experience, allowing long-time drivers to improve their accuracy over time.
For the cardboard truck, our error rates in the later portion of the testing period appeared aberrant from both rain and driving issues, as shown in FIGURE 7. Here, a driver that maintained high accuracy for months early in the testing period suddenly had a string of high error rates over an extended time that were not caused by rain events.
This led me to review the sequence of hardware and software upgrades provided by the on-board metrics vendor. After reviewing the trend lines for accuracy on the cardboard truck, I found that since the software upgrade of late February/early March, error rates were rampant. After discussions with the vendor about the problem, with their guidance I installed both new hardware and software. Immediately, accuracy was reinstated and the system was working wonderfully.
As I have moved into this new frontier of accurately monitoring and reporting individual waste rates, this project took patience and an underlying confidence (faith?) in the technology. It was rewarding to resolve our error issues one by one not only for success of the project as defined, but more importantly, so we may continue to lead the charge in sustainable campus operations.
This article is an abbreviation of the presentation initially given at the California Higher Education Sustainability Conference at UC Santa Barbara on June 24, 2013. I would like to thank Roger Edberg (Senior Superintendent, UCSC Grounds Services), the heavy equipment operators at Grounds Services, and the technicians at Creative Microsystems Inc. (Loadman) for their support and guidance throughout the project.
