Increased performance of a bioreactor using sludge microscopy
Sludge sample collection and upload at any time.
Immediate assessment of the image, indicating the condition of the microcosm in the reactor, and recommendations for process management.
AI algorithms allow correlations between the microscopy images and other process conditions.
Situation
Maintaining the balance between productivity, regulatory compliance, and economy by continuously maximizing the efficiency of the reactions performed by the microbiome present within the biological treatment system is a challenging task. Failure could lead to consent limits being threatened due to solids carryover or incomplete treatment.
Task
To implement microscopy analysis embedded in the routine plant performance monitoring regime combined with an automated, self learning, and continuously improving image analysis tool to provide process guidance.
Action
Using a self-learning computer program to perform feature recognition in microscopic images provides algorithmic precision in selectivity, continuity, and sensitivity in the pattern analysis and recognition.
These results are often more consistent than getting subjective evaluation of microscopy images based on visual observation from different observers, technicians, or consultants.
If such a process is embedded in the automation framework of the plant, then process control variables, such as throughput, dissolved oxygen (DO) levels, recycle activated sludge, and sludge wastage rate can be adjusted using recommendations from such microscopic image analysis.
It is even possible to correlate the microscopy images with the readings of an Oxidation-Reduction Potential (ORP) sensor.
However, embedding an image classification based automated and objective recommendation system for the bioreactor can even help operators with minimal microscopy and image analysis experience to ensure optimal process stability.
The software immediately provides an assessment of the image, indicating the condition of the microcosm in the reactor, and recommendations for process management.
Images from different dates can be compared using the tool to assess trends in the reactor operation.
The tool is also being used as a reinforcement learning platform by the operations team at the plant where they are guiding the AI algorithms regarding the correlations between the microscopic images and other process conditions.
Result
Digital transformation provides water treatment plants unique opportunities for integrating laboratory analysis, water quality, and process monitoring data into operations decision-making and process control.
The tool can be configured readily to provide a tool for uploading microscopy images, catalog these, perform image processing and classification, and utilize machine learning to provide process guidance.
With the tool in place, the operations and engineering team can collect a sludge sample at any time and upload the microscopy image to the software.