Conditioning enables users to summarize or calculate values for raw data streams over a period of time by using the aggregates features, apply a minimum change filter to reduce noise using the deadband feature, or apply logic and state by creating custom conditions.
Although industrial data can be collected at high rates (i.e., millisecond rates), in many cases, the resolution of the data required is much lower. Reducing the volume would result in cost savings and easier handling. The aggregate functionality in HighByte Intelligence Hub takes raw input values and buffers them over a period, then provides calculated values or access to the raw data set to perform any calculation. The out-of-the-box aggregate calculations include min, max, average, delta, and count. Aggregate duration can be defined by length of time (rolling window) or event triggers for on and off.
Sensors with high levels of precision can produce noise that often isn’t needed, can be costly to store, and can cause issues with ML/AI algorithms. Applying a deadband allows the data to be filtered so that noise or minor changes are ignored. The Intelligence Hub centrally manages deadband and can easily apply it to one or multiple inputs.
There are many cases where raw input data must be transformed or conditioned before it is ready for modeling. Maybe it is an old SOAP API that returns escaped XML, or a more advanced case where a user wants to return a raw count from a PLC into an infinite counter. Custom conditions allow users to apply logic and state using JavaScript to a raw input. This allows users to transform arrays of data into objects, check if SQL tables have changed between reads, and much more. Using global functions, users can define JavaScript snippets once and use them in any JavaScript expression applied to a custom condition.
Buffer and perform mathematical computations on raw data
Calculate Min, Max, Average, Delta, and Count over a fixed or sliding window of time
Reduce data changes by applying a minimum change filter
Manipulate and transform raw input data into a usable format for modeling
