Real time forecasting through simulation


Hosni Adra
Process Improvement Manager/Partner
www.createasoft.com

Real time forecasting through simulation


Hosni Adra
Product Manager/Partner
www.createasoft.com

• Simulation is the behavior of a virtual environment through time.
• The success of any simulation project The success of any simulation project requires
  – Correct representation of the model.
  – Correct data input and constraints Correct data input and constraints.
  – Clear goal definition Clear goal definition.

• Traditional simulation tools Traditional simulation tools
  – Based on a simulation engine that is transparent to Based on a simulation engine that is transparent to the user.
  – Most require heavy coding Most require heavy coding.
  – Perform in a Perform in a “build, run, view build, run, view” mode mode.
  – Limited interfaces during the simulation run.
• Dynamic simulation^® tools
  – The simulation engine can interact with the user.
  – Coding requirements vary Coding requirements vary.
  – Perform Perform “build, analyze build, analyze” mode mode.
  – Extensive interfaces to external tools during the Extensive interfaces to external tools during the simulation run.

• Scheduling is a method to organize the sequencing into the operation in order to attain predictable delivery time.
• Forecasting is a look ahead to predict the behavior of the model through time.

• Model definition
  -- Flows, routings, capacity, buffers, material movements, …
• Model variability
  -- Scrap rates, downtime, repair schedules, production sequence, …
• Model constraints
  -- Resources, required equipment, speed, physical constraints, …

• Dynamic data comprises of
  -- Current state data
  -- Current production sequence
  -- Failures, maintenance schedules
• Required additional data
  -- Resource availability
  -- Variation in speeds, buffers, and process requirements.

• Bar coding
  -- Controlled by human factors
  -- Most bar coding sequence do not correctly associate time with location
  -- Can not accurately locate resources
  -- Does not capture motion
  -- Cheap, readily available
  -- Can be found in most environments

• Static RFID rely on passing the RFID chip within proximity of a reader.
  -- Requires the presence of a reader at every scanning station
  -- Is not dependent on the operator
  -- Can correlate time and location within reason
  -- Cheap tags, requires multiple readers
  -- Does not capture motion data

• RFID device emits a signal to a set of receivers which co-locate the position of the tag.
  -- Dynamically updates every 10 sec
  -- Can be applied to resources, products, and material handlers.
  -- Receivers can be located away from processing equipment
  -- Is not affected by human factors
  -- Can be applied to locate based on floor level (Patent Pending)
  

• Constraints can be acquired through
  -- ERP/MRP systems
  -- Input from the processing floor, or equipment
  -- Include failures, scrap rates, QA rates, rework
• This set can be ignored but will affect the results of the forecasted model

• Virtual Camera
  -- Live view of the operation (based on the tracking solution)
• Historical replay of events
• Constant forecasting of future events
• Accurate costing analysis
• Improved scheduling capability

• Solution is being implemented and tested at a large warehousing facility
  -- Utilizing active RFID tags
  -- Using Simcad Pro as the forecasting engine
  -- Connects to an ordering system for future pulls.
  -- Tracks parts and resources

• Tracking is only visual unless
  -- Resources, parts, and equipment are mapped based on relationships
     - Resources can be within proximity of a piece of equipment, with the presence of a part indicates that the resource is working on the part (with limitation)
     - Resource moving within proximity of a forklift for the duration of the move indicates that the resource is tagged to the truck.

Download this presentation

• Active RFID tags are accurate within 10 sq ft.
  -- Inaccuracy is also generated by the type of building.
  -- Signals from other devices may create interference.
  -- Solution – utilize additional receivers with smart positioning to eliminate wrong signals.

• Proprietary locator engine that feeds off the main RFID data.
  -- Locate using multiple receivers (3 or more)
  -- Locate height and planar location
  -- Connects to Simcad Pro for dynamic data feed through SQL engine
  -- Automatically creates possible relationships.
  -- Accurate to 3 Sq Ft
  

• The simulation model contains variability due to
  -- statistical distributions
  -- Scrap, rework, and other factors
• To Forecast accurately
  -- Monte Carlo simulation is used to make at least 100 runs of the model before a notification event is sent.
  -- Monte Carlo results are combined to create the most probable future scenario
  

• Can be applied to any process flow.
• A combination of active and static RFID tags can be used to reduce implementation cost.
• Not limited to a specific RFID vendor.
• Requires a dynamic simulation^® environment to operate properly.
  

• Example
• Q&A