Smart Buildings and Contaminant Events Benchmarks

 

The dispersion of contaminants from sources (events) inside a building can compromise the indoor air quality and influence the occupants' comfort, health, productivity and safety. Such events could be the result of an accident, faulty equipment or a planned attack. Under these safety-critical conditions, immediate event detection should be guaranteed and the proper actions should be taken to ensure the safety of the people. In this paper, we consider an event as a fault in the process that disturbs the normal system operation. This places the problem of contaminant event monitoring in the fault diagnosis framework of detection and isolation.

In the following you can find the datasets generated by the Matlab-CONTAM Toolbox for Contaminant Event Monitoring in Intelligent Buildings that you can find in the i-Sense open library section under the Creative Commons Attribution-NonCommercial-ShareAlike 3.0 open License.

 

References

 

  • Michaelides, M. P., Reppa, V., Christodoulou, M., Panayiotou, C.G., Polycarpou, M.M., (2013) “Contaminant Event Monitoring in Multi-zone Buildings Using the State-space Method”, Building and Environment, vol. 71, Pages 140-152, Jan. 2014 ref

  • Michaelides, M. P., Eliades, D. G., Christodoulou, M., Kyriakou, M., Panayiotou, C., & Polycarpou, M. (2013). A Matlab-CONTAM Toolbox for Contaminant Event Monitoring in Intelligent Buildings”, in Proceedings of LEAPS, the 1st Workshop on Learning strategies and data processing in nonstationary environments, Artificial Intelligence Applications and Innovations Conference (AIAI 2013), Paphos, Cyprus, pp. 605-614 ref

  • Michaelides, M., Reppa, V., Panayiotou, C., Polycarpou, M.: Contaminant event monitoring in intelligent buildings using a multi-zone formulation. In: Proc. of SAFEPROCESS. Mexico City, Mexico (2012) ref
  • Eliades, D., Michaelides, M., Panayiotou, C., Polycarpou, M.: Security-oriented sensor placement in intelligent buildings. Building and Environment 63, 114--121 (2013) ref
  • Wang, L., Dols, W., Chen, Q.: Using CFD capabilities of CONTAM 3.0 for simulating air flow and contaminant transport in and around buildings. HVAC&R Research, 16(6), 749--763 (Nov 2010) ref

Smart Building Datasets with Contaminant Event


We use the Holmes House building scenario. We assume a single source of the contaminant of interest (i.e. CO2) that is released at time 1 hour and the simulation lasts for 24 hours. Furthermore, we assume sensors in all 14 of the building zones.

Download the Smart Building Datasets with Contaminant Event Benchmark

Objectives

  • Detection of contaminant event
  • Isolation and identification of contaminant event

 

Dataset Description

 Y -> Data Set 1: Single Contaminant Source at a constant rate 100 g/h in wind directions [0, 90, 180, 270]O and with 14 different (zones) source locations. The release of contaminant source start after 1 hour.  Simulation Time: 24 hours.
Y1 -> Data Set 2: Variable Single Contaminant Source with step every 2 hours from 50 to 100 g/h in wind directions [0, 90, 180, 270]O and with 14 different (zones) source locations.
Y2 -> Data Set 3:  The same with Y1 but with initial concentration in zones with 20 g.
Y3 -> Data Set 4: Variable Single Contaminant Source with function 100*|sin(2πt)| in wind directions [0, 90, 180, 270]O and with 14 different (zones) source locations.
Y4 -> Data Set 5: Contaminant Source from outside at a constant rate 50 g/h in wind directions [0, 90, 180, 270]O.
Y5 -> Data Set 6: Single Contaminant Source into zone 7 at a constant rate 100 g/h with wind direction 90O.  Also, with combinations [Close HalfOpen Open ] on paths opening in zone 8 [P13 P15 P17 P18 P22]
Y6 -> Data Set 5:  The same with Y5 but Contaminant Source location into zone 14.

 

Smart Building Datasets with Sensor Fault


We use the Holmes House building scenario. The contaminant of interest (i.e. CO2) is present in the atmosphere with a mean concentration of 50 mg/m^3/sec modeled as a pseudorandom sequence. The time interval between the transitions (jumps) follows a Markov process with 0.1 transition probability while the magnitude of the sequence after a transition is a random number between [45, 55]. For the simulations, we assume wind direction 90deg (from the east), wind speed 10m/s and fully open leakage path openings. Furthermore, we assume sensors in all 14 of the building zones. 
Scenario description:
The sensors are monitoring the contaminant of interest and a single, permanent, additive fault is introduced at time t=2000 sample in the sensor placed in the kitchen (Z5).

Download the Smart Building Datasets with Sensor Fault Benchamark

 

Objectives

  • Detection of sensor fault
  • Isolation and identification of sensor fault
  • Data reconstruction for faulty sensor data

 

Dataset Description


Each dataset is a <.mat> file that loads the structure Y. The structure contains the following entries:

  • Y.Time (1x10000) = Time vector with time samples every 0.0167 sec.
  • Y.Amartices (14x14) = State transition matrix corresponding to the specific experiment
  • Y.Q (14x14) = Flows matrix
  • Y.Qext (14x1) = External flows vector
  • Y.Concentration (14x10000) = Time series of concentrations in the building zones without any faults
  • Y.Source (14x10000) = Time series of external concentrations created by outside source
  • Y.Sfault_... (14x10000) = Time series of concentrations in the building zones with permanent, additive sensor fault introduced at t=2000 in the Z5 sensor placed in the kitchen.
    • …Z5freeze = sensor fault frozen at value 50
    • …Z5offset = offset fault of magnitude 10
    • …Z5drift = drift fault with slope .01
    • …Z5noise = noise fault with Gaussian white noise of variance 10

For example, to plot the contaminant concentration in the kitchen for the first dataset for the frozen fault case, I would use the following commands from Matlab:
>> load Ysfault
>> plot(Y.Time, Y.Sfault_Z5freeze(5,:))

 

Smart Building Datasets with Contaminant Event and Sensor Fault


We use the Holmes House building scenario. The contaminant of interest (i.e. CO2) is present in the atmosphere with a mean concentration of 50 mg/m^3/sec modeled as a pseudorandom sequence. The time interval between the transitions (jumps) follows a Markov process with 0.1 transition probability while the magnitude of the sequence after a transition is a random number between [45, 55]. For the simulations, we assume wind direction 90deg (from the east), wind speed 10m/s and fully open leakage path openings. Furthermore, we assume sensors in all 14 of the building zones. 

Download the Smart Building Datasets with Contaminant Event and Sensor Fault Benchmark


Scenario description
The sensors are monitoring the contaminant of interest and a single, permanent, additive fault is introduced at time t=2000 sample in the sensor placed in the kitchen (Z5). In addition to the sensor fault, a contaminant source of emission rate 100 mg/s is also placed in the kitchen (Z5) at time t=2000 sample.

Objectives

  • Detection of sensor fault
  • Isolation and identification of sensor fault
  • Data reconstruction for faulty sensor data
  • Detection of contaminant event
  • Isolation and identification of contaminant event

 

Dataset Description

Each dataset is a <.mat> file that loads the structure Y. The structure contains the following entries:

  • Y.Time (1x10000) = Time vector with time samples every 0.0167 sec.
  • Y.Amartices (14x14) = State transition matrix corresponding to the specific experiment
  • Y.Q (14x14) = Flows matrix
  • Y.Qext (14x1) = External flows vector
  • Y.Concentration (14x10000) = Time series of concentrations in the building zones without any faults
  • Y.Source (14x10000) = Time series of external concentrations created by outside source
  • Y.Sfault_... (14x10000) = Time series of concentrations in the building zones with permanent, additive sensor fault introduced at t=2000 in the Z5 sensor placed in the kitchen.
    • …Z5freeze = sensor fault frozen at value 50
    • …Z5offset = offset fault of magnitude 10
    • …Z5drift = drift fault with slope .01
    • …Z5noise = noise fault with Gaussian white noise of variance 10

For example, to plot the contaminant concentration in the kitchen for the frozen fault case, I would use the following commands from Matlab:
>> load Ysfault2
>> plot(Y.Time, Y.Sfault_Z5freeze(5,:))

 

 

 

 

We are committed to maintaining a public repository of i-Sense benchmarks in the spirit of cooperative scientific progress as promoted by digital open access guidelines to EU funded research oriented towards the dissemination of results and outcomes.

You are free to download a portion of the datasets for non-commercial research and educational purposes.  Work based on the dataset should cite our references papers listed with each benchmark.

All datasets and benchmarks on this page are published under the Creative Commons Attribution-NonCommercial-ShareAlike 3.0 License.

Creative Commons License