Definition
A Virtual Sensor estimates product properties or process conditions using mathematical models. These mathematical models use other physical sensor readings to calculate the estimated property or condition.
When Do You Need Virtual Sensors?
Virtual sensors are useful, often necessary, when:
The Property or State Cannot be Measured by a Physical Device
Measuring some properties can be quite complicated and may involve significant processing of product samples by hand, may include shelf aging time, humidity tests, tensile strength measures, burning the product and measuring the ash, many of the reasons that QC labs exist.
A Physical Sensor is Too Slow
Some sensors take time to cycle, like gas chromatographs. The sensor's or instrument's cycle time may be slower than what is needed for process control or feedback to the operators. A Virtual Sensor can provide information between readings. Our "tracking" virtual sensors can optionally self-calibrate against such instruments, leading to high on-going perpetual accuracy.
A Physical Sensor is Too Far Downstream
A physical sensor may be too far downstream, physically, to give timely information for process adjustments. This lag time can hurt good control and operator response. This is like turning the wheel of your car and having the wheels respond some minutes later. You are likely to have difficulty steering. Virtual sensors can be predictive, creating an estimate now for a reading that will occur in the future. This is a common application for virtual sensors.
Implementing a Physical Sensor is Too Expensive
Some sensors are REALLY expensive, costing 10's of thousands or even 100's of thousands of dollars. If the sensor is very expensive you might be able to afford just one and find ways to channel production from several units or lines through it from time to time. If this is the only way to measure the property, you can use these measures to estimate the property on-line for all production units, virtually multiplying your physical sensor investment. For example, many oil companies have one production meter per platform and put each well "on test" to measure production. In this case, virtual flow meters can be applied to all wells using this data.
There is No Means to Install a Physical Sensor
There is sometimes just no room to put in the sensor, or the geometry causes the sensor to be inaccurate. The sensor may require more space than is available.
The Sensor Environment is Too Hostile
There are processes where measuring the property can only be done intermittently because the process is so hostile that it damages the sensor. Virtual sensors can use the data collected occasionally to continuously "see" into these environments without a physical presence.
A Physical Sensor is Inaccurate (Drifts)
Some types of sensors struggle to maintain calibration, due to their design or operating environment. You can build a virtual sensor using data when the sensor is freshly calibrated and put it on-line to beat the accuracy of the physical sensor.
A Physical Sensor is Expensive to Maintain
Some sensors get damaged by the process and have to have regular PM's performed to keep them operating. One example is a positive displacement meter on an oil well, processing some amount of sand. The meter erodes and loses accuracy and if on a platform, a barge has to be ordered and a crew dispatched to replace the meter so that it can be taken to the maintenance shop or sent to the vendor to be rebuilt. A virtual meter does not suffer from sand in its gears.
How Do You Build and Implement Virtual Sensors?
You gather data that relate to the predicted result. You then build and validate a mathematical model of the relationships using Expert, our data modeling software. You then place the model on-line using our Intellect server software. Intellect supports unlimited sensor implementations and can estimate properties as often as you need; every so many hours, minutes or even sub-second, up to the performance limitations of the computer.