Costa Mesa Sanitary District: Using ML to reduce manhole maintenance cost and boost safety

Reduced manhole maintenance cost by 60%


Safer, smoother streets for Costa Mesa and more efficient manhole maintenance using machine learning

The Costa Mesa Sanitary District (CMSD) is dedicated to protecting public health and the environment for current and future generations. The district, which serves more than 118,000 residents, provides solid waste and wastewater collection services to Costa Mesa, California, as well as parts of Newport Beach and unincorporated Orange County.

Business Situation

Tackling a hidden–but vital–infrastructure maintenance challenge

Manholes are everywhere, but it’s easy to overlook how important they are. These omnipresent portals provide critical access to underground public utilities for inspection, maintenance, and upgrades. Failure to maintain them properly can cause numerous problems, from road hazards to sewer blockages and delays in repairing utilities.

Costa Mesa Sanitary District (CMSD), traditionally spent a significant amount of time and money—more than $100,000 annually—maintaining manholes. The process required many staff hours to manually survey, rate, and service approximately 5,000 manholes over 218 street miles.

To save time, make the most of taxpayer resources, and elevate overall maintenance efforts and public safety, CMSD wanted to streamline manhole maintenance with a technology-forward and scalable solution.

Costa Mesa Sanitary District

“An intangible benefit is safety. CMSD employees are not subjecting themselves to potential life-threatening injuries by being struck from fast moving vehicles when they were in the street manually inspecting manhole covers.”

Scott Carroll, CSDM, ICMA-CM, General Manager, Costa Mesa Sanitary District


Leveraging ML to drive new insight, and maintenance efficiency

CMSD collaborated with Google Cloud and SpringML to create a solution that streamlines manhole maintenance using a GoPro camera, Google Cloud technology, and the power of machine learning (ML) to detect sewer manholes, analyze them, and rate their conditions.

Every quarter, a CMSD team member drives a car outfitted with a GoPro camera through the entire district of over 218 street miles, which includes the city of Costa Mesa and small portions of Newport Beach, to detect all manholes. Since privacy is a top priority, the system limits image processing to the section of road in front of the vehicle.

After each day of recording, the driver uploads images and videos from the GoPro SD card to a local server.

The solution then automatically loads the data into Google Cloud Storage. A Google Cloud Scheduler workflow detects any new videos at the end of each day and uses machine learning to review the images and videos, identify any type of damage, and rates each manhole’s condition on a 1-5 scale. CMSD stores final scores in Google BigQuery.

CMSD team members access results through a web application that identifies which manholes require maintenance. Two staff members then review the results and determine repair priorities.

Costa Mesa Sanitary District

“Another benefit is the ability of transferring nearly 700 work hours employees were spending a year to manually inspecting manhole covers to other operational areas such as spending more time cleaning the sewer system and/or performing closed circuit televising of pipeline to evaluate the current condition of the asset.”

– Scott Carroll, CSDM, ICMA-CM, General Manager, Costa Mesa Sanitary District

The solution continues to learn and improve based on feedback submitted via the web application. For example, if the model inaccurately detects a manhole, a CMSD team member can mark that in the web application, and the solution uses their feedback to refine the model.


Optimizing resources, reducing costs, and improving public safety with a single cloud-based ML solution

Since launching the initiative in October 2020, CMSD has realized immediate and sustained improvements to its maintenance program. It automated a significant portion of the manhole maintenance process, saving significant labor time and taxpayer resources.

CMSD cut costs associated with manhole maintenance by 60 percent. And the assessment process now only requires one driver and car to patrol streets and two staff members to review results – freeing significant resources for other projects and priorities.

The initiative is also helping to improve public safety with more frequent and effective manhole maintenance, resulting in smoother roads, avoiding sewer blockages, and improving worker access to underground public utilities.

SpringML Logo

What we’re doing in CMSD is helping them innovate with modern technology to create better outcomes. It’s a framework we’ve been working on to come up with an end to end solution that turns existing fleet vehicles into intelligent sensors. I’m thankful for for Scott’s leadership to explore new technology and showing others in his industry what’s possible.

Eric Clark, Vice President, Public Sector, SpringML

CMSD’s manhole maintenance project shows how leveraging ML can improve the effectiveness of a necessary local government function while saving money and labor. Not only does this scalable solution streamline manhole maintenance, it also allows more frequent reviews of manhole conditions and provides a historical view of how the district’s manholes change over time.

Thought Leadership

Let's chat