Fault Detection

Fault Detection

Stop the squander! Identifying building malfunctions that waste energy

Designing New Methods for a Fault Detection and Diagnosis in Buildings


What would happen to the electricity bill of a particular building if heating, ventilation, or air conditioning systems are left running when they shouldn’t?  For example, what would happen if the air conditioner runs on a cold day?  If your classroom is in that building, you would be cold and wondering if the system is broken. Perhaps someone brings in space heaters because everyone is so cold. Something in the building is working improperly, or faulting.

Building “faults” occur in all types of buildings. Faults in commercial buildings, which include schools, stores, offices, churches, and restaurants among others, waste an estimated 1 quadrillion BTU of primary energy 1 annually. While these faults are often difficult to identify, this amount of energy is equal to more than 65% of the total energy consumption of Colorado.

Consider the following case as an example. A rural middle school owns and operates a 9,300 sq. ft. building and has almost no budget for maintenance. In November 2017, the director noticed an increased utility bill, which was typically 65-75kWh/day. By March 2018 the energy use was 150 kWh/day, but the director could not figure out what was causing the uptick. The director began shutting off different mechanical systems one at time and monitoring the electricity meter. Students got involved and helped collect and analyze data, but even in June 2018, the director still wasn’t certain about what part of the heating, ventilation, and air conditioning (HVAC) was faulting. 

Scenarios like this one present the need for solutions that can automatically detect when a building isn’t working improperly (i.e. faulted) and identify the root cause (i.e. diagnose). This process is called fault detection and diagnostics.

The Challenge

The Challenge objective is to develop new methods to accurately and efficiently identify when faults occur and the type of fault based on different levels of available information from the building; and to develop a straightforward system of notifying and visualizing the fault and its impact for building owners. These could be addressed using new algorithms, processes/flowcharts, and visual interfaces.

For this Challenge we will use real data from an actual building, a 1,600 square foot small office building constructed for the purpose of conducting building experiments, the Flexible Research Platform (FRP). The system in the FRP is a multi-zone HVAC system with 10 conditioned rooms that can be controlled individually. The type of HVAC system is a conventional rooftop unit (RTU) variable air volume (VAV) system with terminal reheating. Faults can be intentionally applied and resulting data recorded on whole building energy use, sub-metered end energy use, and indoor temperatures and humidity. Examples of faults tested in the FRP include supply/return duct leakage, non-standard refrigerant charge, condenser fouling, HVAC setback errors, and thermostat faults.    

For this Challenge, there are three sets of data from the FRP:

  • Only whole building energy use
  • Sub-metered energy use that includes building lighting, plug load, HVAC fan, compressor, reheating, etc., and
  • Data with additional zone thermostat setpoint and actual zone temperature.

See the More Details section below to access the files and an introductory presentation on the Challenge. The data includes building/HVAC operation data both for baseline (i.e., normal operation without faults) and fault events. The type of possible faults is also provided with the data sets.

In formulating your idea submission, JUMP into STEM seeks your ideas on the following:

  • An identification of what type of faults and when they occur;
  • A method for indicating how to identify the fault per different level of available data;
  • Energy savings (kWh)potential by fixing the identified faults; and
  • A method that will help building owners visualize the fault (through data or other method).

For information on how the JUMP into STEM process works, please visit the How it Works section.

What Idea Responses Should Include

Ideas can be up to 1,500 words and may be attached as a PDF in the JUMP into STEM response form. Winning submissions should be thoughtful, well-articulated, innovative and/or unique.  Be sure that your responses describe 1) what are the types of faults found and when they occur, 2) the new methodology to detect the fault event, 3) how the energy savings can be estimated by comparing energy use during the fault event vs. normal operation, and 4)a method that will help building owners visualize the fault(s).

A few guidelines to keep in mind:

  • All are welcome: All community members are welcome and encouraged to participate in the dialogue.
  • Be respectful: Please, no remarks that are off topic or offensive.
  • No solicitation: Please, no promotions or endorsements for specific commercial services or products.
  • Response time: Where applicable, a JUMP team member will respond to process and program related questions within 2 business days.

Active JUMP into STEM Challenges will be open for idea submission, voting, and commenting for a minimum of 8 weeks. A panel of judges will select the Finalist Awards based on the idea’s potential impact on reducing energy consumption in buildings as well as a review of how unique, innovative, well-articulated and thoughtful the idea submission is. See the Schedule section for specific due dates and milestones. 

One of the unique advantages of the JUMP crowdsourcing community is that it engages users in evaluating and reviewing the ideas through the "voting" and "comment" functions. The ORNL,NREL, DOE, and University Partners use this information to gauge interest in the topic and idea submission. A leading number of votes or comments, however, does not guarantee advancement, but is one indicator in the judging phase of considering ideas for Finalist Awards.

For More Information

Be sure to check out the Round 2 webinar series – a creative virtual seminar forum where you can learn more about the Round Two Challenge and industry’s experiences with fault detection and diagnostics.

Webinar 1: Designing Methods for Understanding Building Faults, Including an Examination of the Challenge Sample Data Sets with Piljae Im, PhD, -- a researcher with the Building Envelope & Urban Systems Research (BE&USR) Group in the Building Technologies and Research Integration Center (BTRIC) at ORNL.
Date:  October 24, 2pm to 3pm ET

Webinar 2: Round 2 Challenge Overview and an Introduction to Building Performance Fault Detection and Diagnosis with Stephen Frank, PhD, Senior Systems Engineer with the National Renewable Energy Laboratory

Date:  October 26, 1pm to 2pm ET

Webinar 3: Building Owners and Manager’s Perspectives on Addressing Building Performance with Jared Parker, building manager with Michigan State University Facilities Management
Date:  October 30, 2pm to 3pm ET

More Details

Data Files

1: Energy from raw fuel, so when calculated, takes into consideration electricity distribution losses and power plant efficiencies.