Hitachi, Ltd., new function utilizing AI for water and sewage business related cloud

Hitachi has added new functions for equipment diagnosis, water quality prediction, and operation support using artificial intelligence (AI) to its lineup of “O & M (Operation & Maintenance) support digital solutions”, a cloud service that supports water and sewage services. .. The company has been offering this solution since October 2018. The functions expanded this time are the “equipment condition diagnosis function” that enables condition-based maintenance by diagnosing the condition of equipment such as pumps and blowers, and the “water quality” that predicts the quality of raw water and supports the optimization of chemical injection amount. It becomes a “prediction function” and a “plant operation support function” that proposes future demand forecasts and operation plans based on the learned operator’s know-how and judgment. Condition-based maintenance is condition-based maintenance, and refers to the concept of determining the necessity of maintenance in consideration of the deterioration status and failure risk of equipment, and performing maintenance before failure or usage limit. In addition, we will expand our data analysis services by adding a “data visualization function” that calculates and displays useful indicators for operation management and maintenance by combining monitoring and inspection data collected in the cloud. Demonstration experiments have been conducted at water purification plants commissioned by Hitachi to confirm the effectiveness of each of these functions.

Conceptual diagram of “equipment condition diagnosis function” The equipment condition diagnosis function collects operation data of equipment such as pumps and blowers, and uses the ART (adaptive resonance theory) method, which is a kind of data clustering technology of AI, to normalize the past. Pre-learn the operation data of various equipment. As a result, the correlation of data that serves as a reference for predictive diagnosis is classified, and a category of normal data is automatically generated. Then, new data acquired during actual equipment operation is automatically classified and compared with the normal category to diagnose whether the operating condition is normal. As a result, condition-based maintenance that catches state changes such as defects at an early stage becomes possible, which contributes to the reduction of loss costs due to large-scale failures and the reduction of maintenance costs by extending the maintenance interval.

Conceptual diagram of “water quality prediction function” The water quality prediction function uses deep learning technology to build a prediction model using past operation record data and open data such as environmental conditions (weather and water source) as training data, and the prediction results are sequentially generated. Present, predict the condition of raw water and the quality of treated water several hours ahead. By being able to support the setting of objective and appropriate operating conditions to obtain good water quality, we will standardize operations and manage water quality without depending on the know-how and judgment of experts.

Conceptual diagram of “plant operation function” The plant operation support function learns know-how and judgments such as operator knowledge and equipment operation conditions by using reinforcement learning technology and statistical analysis, which are a type of AI, and the future is based on the learned conditions. Support the work of operators by proposing demand forecasts and operation plans. This makes it possible to support the standardization of operations and the transfer of know-how and skills of skilled workers. In addition, the residual salt management function will be available in 2021. This function calculates the recommended value of chlorine agent injection rate to satisfy the target value of residual salt in distribution reservoirs and faucets using the reaction model and data such as chlorine injection rate and water temperature, and reflects it in the plant operating conditions. By doing so, we will support a stable supply with appropriate water quality.