Transcription

Case Study: Video Analytics OilMist DetectionthFike Corporation, 704, SW 10 Street, Blue Springs, MO 64015 U.S.A.Fike.comCASE STUDY

Video Analytics Oil Mist DetectionIntroductionVideo Analytics for Engine RoomsFike Video Analytics Server detects flame, smoke and oil mist by processing video from off the shelf IPcameras in areas such as machinery spaces or mooring decks.The Fike Server is installed and connected to the existing camera LAN, detection event alarms and livevideo may be monitored in the engine control room and elsewhere, sent digitally to the VideoManagement System (VMS), Safety Management and Control System (SMCS) or viewed through the FikeSpyderGuard VMS.Software alarm/exclusion zones and alarm verification functionality, combined with Fike’s uniquealgorithm sensitivities, allow for endless configuration options to suit specific fire, smoke and oil mistdetection requirements.There are three, mutually exclusive analytics algorithms used; smoke, flame and reflected flamedetection. The smoke detection algorithm is used to detect smoke, oil/fuel mist or spray from a leakunder pressure and oil vapor that is generated when fuel or oil leaks onto a hot surface.Oil mist may form when high pressure fuel oil, lubricating oil, hydraulic oil, or other oil is sprayedthrough a narrow crack, or when leaked oil connects with a high temperature surface, vaporizes andcomes in contact with low air temperature.When the concentration of oil mist increases and reaches the lowest explosion level (LEL; 50 mg/ℓ, asdefined by the IACS), explosion may occur when the mist contacts surfaces of over 200 C (392 F) or aspark.CASE STUDY1

Video Analytics Oil Mist DetectionExample System SetupThe following is an example of a system on board a cruise ship fitted with IP cameras and monitored byan existing video management system.Fike Video Analytics (FVA) Servers are installed in the CCTV server racks and connected to the cameraLAN. The Fike servers are capable of analyzing video from 16 cameras and have 6TB of recording space,providing average 2 weeks of recording time.The video management system client software is running on a PC in the Engine Control Room andbroadcasting live video on four, 40” CCTV screen monitors. The FVA management system, SpyderGuard,was installed on the PC as a means for configuration management of the analytics and, in the event ofdetection, pops-up live video, with analytics overlay, to one of the large CCTV monitors. Audible andvisual alarms are also provided.Prior to installing a video analytics detection system, a hazard analysis is performed to identify areas fordetection, covering engines, purifiers, thermal oil heaters and other equipment involving flammableliquids.With these areas in mind, camera location is best if installed with a tilt of approximately 45 degree anglelooking down and approximately 20ft from the equipment. When the camera is installed and connectedto the server, the analytics settings can be modified to target specific areas for highest sensitivity.Smoke (oil mist) alarm zones are configured above the equipment so that alarm will occur inside thezone. The analytics track the detection through the entire field of view, however will not alarm untildetection reaches alarm zone.Oil mist is hot and under high pressure, therefore rises, and airflow in the spaces also works to move themist into the zones.Areas with low potential for origination of smoke, mist or flame, such as the walkways will be excludedfrom the detect zones to avoid unwanted detections.CASE STUDY2

Video Analytics Oil Mist DetectionFike Video AnalyticsOil Mist/Fire DetectionVideo Management SystemSystem ECR MonitorsAnalytics OverlayPop-up on existingVMS monitorVMS Client WorkstationRunning SpyderGuard –Fike VMSCamera DisplayAlarm/Trouble IPrelay to ship’sautomationmonitoring systemUSBAlarm LightVMS Servers3CCTV LANFike ServersCCCCCCIP CamerasSystem LayoutCASE STUDY

Video Analytics Oil Mist DetectionThe analytics identify areas of light contrast in motion and movement across pixels, the outline of aperson close to the camera can cause false detections. Time delays, sensitivity and alarm zoneadjustments are made while monitoring normal operation of the crew.By configuring alarm zones throughout the field of view, avoiding areas where personnel are normallypresent, and smoke (oil mist) detection sensitivity can be set to High. High sensitivity allows detectionand tracking of minute mist particles, often invisible to human monitoring.Oil mist is hot and/or under high pressure, therefore rises, and airflow in the spaces also works to movethe mist into the 3-dimensional zones.Flame and reflected flame detection are configured to Medium sensitivity as they are enabled in theentire field of view and a brief delay is added to avoid transient events that could cause false detections.Below are screenshots with smoke alarm zones highlighted within the blue frames.4CASE STUDY

Video Analytics Oil Mist DetectionThe following are two images indicating targeted zoning with dark blue lines and smoke detections atengine startup. The light blue outline is one smoke detection algorithm display, there are also light bluetiles representing detection from another algorithm, barely visible, well into the zones.High sensitivity smoke detection is enabled in the entire camera field of view. Smoke is detectedanywhere in the field of view and tracked. When smoke breaks the plane of the zone, there is an alarmimmediately and the detection indication expands to all areas where there is visible smoke.5CASE STUDY

Video Analytics Oil Mist DetectionWhen the water mist fire suppression systems were tested, the mist migrated to adjacent engines. Theanalytics detected the mist that looked very similar to oil mist or smoke; 8 cameras detected mist duringone of the tests. Below are pictures of the water mist detections.6Water mist detectionsCASE STUDY

Video Analytics Oil Mist Detection7Above is water mist detection, smoke type detections are represented either by a blue outline or bluetransparent tiles overlay. Light smoke can be generated when engines are first started. The followingimages are detections from these events.CASE STUDY

Video Analytics Oil Mist DetectionEngine startup smoke8Engine startup smokeCASE STUDY

Video Analytics Oil Mist DetectionAbove are screenshots during tests in machinery spaces using a calibrated smoke emitter. Althoughsmoke detectors were directly above the smoke emitters, none of the existing detectors went intoalarm.9Flame from welding was detectedCASE STUDY

Video Analytics Oil Mist DetectionSmoke emitter tests were also conducted on mooring decks10SpyderGuard BrowserCASE STUDY

Video Analytics Oil Mist DetectionSpyderGuard TimelineFeatures of SpyderGuard Video Management System View live video with analytics overlay of detectionConfigure analytics settingsView and download videos of previous detectionsProvides alarm light and voice notification of alarmSmoke detection Maintenance Mode11A number of SpyderGuard software modifications were accomplished by the Fike software engineers based onfeedback from the engineering crew. These were primarily related to annunciation and acknowledgement ofalarms.One significant new feature to reduce human interference detections is the Maintenance Mode for use whenpersonnel are working directly on monitored equipment. When turned ON, smoke detection on a camera isdisabled for an amount of time designated by the crew. Flame and reflected flame continue to operate normally.Information Provided By:Rick Jeffress ARM, CET, CFPS - Fike [email protected]: USA (443) 562-7404 Copyright 2016, Fike Corporation. All rights reserved. Form No. FPGCS 2016-002CASE STUDY

Smoke (oil mist) alarm zones are configured above the equipment so that alarm will occur inside the zone. The analytics track the detection through the entire field of view, however will not alarm until detection reaches alarm zone. Oil mist is hot and under high pressure, therefor