# US 2024/0004355 A1 · System and Method for Building and Deploying Prescriptive Analytics to Predict and Control End Product Quality in Batch Production Monitoring and Optimization

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2024/0004355 A1
- Document ID: US20240004355A1
- Status: Published application
- Kind: A1
- Application number: 17/809,640
- Filed: 2022-06-29
- Published: 2024-01-04
- Earliest claimed: 2022-06-29
- Assignees: AspenTech Corporation
- Applicants: AspenTech Corporation
- Inventors: Hong Zhao; Ashok Rao; Mir Khan
- Web page: https://patentgrep.com/patent/US20240004355A1

## CPC

- G05B 13/048: {using a predictor}
- G05B 23/024: {Quantitative history assessment, e.g. mathematical relationships between available data; Functions therefor; Principal component analysis [PCA]; Partial least square [PLS]; Statistical classifiers, e.g. Bayesian networks, linear regression or correlation analysis; Neural networks}
- G05B 13/0265: {the criterion being a learning criterion}
- G05B 23/0221: {Preprocessing measurements, e.g. data collection rate adjustment; Standardization of measurements; Time series or signal analysis, e.g. frequency analysis or wavelets; Trustworthiness of measurements; Indexes therefor; Measurements using easily measured parameters to estimate parameters difficult to measure; Virtual sensor creation; De-noising; Sensor fusion; Unconventional preprocessing inherently present in specific fault detection methods like PCA-based methods}
- G05B 23/0275: {Fault isolation and identification, e.g. classify fault; estimate cause or root of failure}
- G06N 20/10: using kernel methods, e.g. support vector machines [SVM]
- G05B 2219/32077: Batch control system
- G05B: CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS (systems for controlling or regulating non-electric variables G05D; systems for regulating electric or magnetic variables G05F; control devices or systems insofar as characterised by mechanical features only G05G)
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS

## Abstract

Embodiments control and optimize batch processes. An embodiment obtains and standardizes historical operating data from a plurality of batch production runs of an industrial process. For each batch production run, the standardized operating data corresponding to the batch is partitioned into one or more stages and one or more signature for each stage is determined using the partitioned standardized data. Each determined signature is associated with a class label based upon whether output of a batch run corresponding to the signature conforms or does not conform with operational standards. A model is trained, with at least a subset of the signatures as inputs and associated class labels as outputs, to predict, based on operating data from a real-world batch process, whether output of the process will conform or not conform with the operational standards. Online predictions can be automatically or manually applied to control and optimize a batch production run.

## Claims

### Claim 1

A computer-implemented method for creating a machine learning predictive model for real-world batch production industrial process monitoring and optimization, the method comprising:
obtaining historical operating data from a plurality of batch production runs of an industrial process;
standardizing the obtained historical operating data for each run of the plurality of batch production runs;
for each batch production run of the plurality: (i) partitioning standardized operating data corresponding to the batch production run into one or more stages and (ii) determining one or more signature for each of the one or more stages using the partitioned standardized operating data corresponding to the one or more stages;
associating each determined signature with a class label based upon output of a batch production run of the plurality corresponding to the determined signature conforming with operational standards or not conforming with the operational standards;
training a machine learning predictive pattern model with at least a subset of the determined signatures as inputs and associated class labels as outputs, wherein the training configures the model to predict, based on operating data from a real-world batch production process, whether output of the real-world batch production process will conform or not conform with the operational standards; and
wherein the obtaining, standardizing, partitioning, determining, associating, and training are automatically implemented by one or more processors.

### Claim 2 (depends on 1)

The method of claim 1 wherein, for each batch production run of the plurality, the obtained historical operating data is configured to include: (i) batch initialization data, (ii) batch progress data, and (iii) batch end product quality data.

### Claim 3 (depends on 1)

The method of claim 1 wherein standardizing the obtained historical operating data for each run of the plurality comprises at least one of:
aligning the obtained historical operating data for each batch production run of the plurality with a specified standard reference batch over time;
detecting and removing outlier runs from the obtained historical operating data; and
labeling each batch production run of the plurality with a class label, wherein each class label indicates whether output of the batch production run conforms with the operational standards or is not conforming with the operational standards.

### Claim 4 (depends on 1)

The method of claim 1 wherein determining one or more signature for each of the one or more stages using the partitioned standardized operating data comprises:
generating one or more engineering features (EFs) or Key Performance Indicators (KPIs), using the partitioned standardized operating data corresponding to the one or more stages; and
grouping the generated one or more EFs or KPIs into a set to form a given signature.

### Claim 5 (depends on 4)

The method of claim 4 wherein the set is an expandable set and grouping the generated one or more EFs or KPIs into the expandable set to form the given signature comprises:
(i) adding the one or more EFs or KPIs into the expandable set over time as the one or more EFs or KPIs are generated; and
(ii) assigning a weight to each of the generated one or more EFs or KPIs in the expandable set.

### Claim 6 (depends on 5)

The method of claim 5 wherein the expandable set increases in size over the time with progress of a given batch production run.

### Claim 7 (depends on 1)

The method of claim 1 further comprising:
receiving input indicating a selected signature type; and
wherein, in determining a given signature, the selected signature type is determined.

### Claim 8 (depends on 1)

The method of claim 1 wherein each signature is determined at pre-defined points of each stage.

### Claim 9 (depends on 1)

The method of claim 1 wherein the obtained historical operating data is collected from one or more sensor measurements.

### Claim 10 (depends on 1)

The method of claim 1 wherein training the machine learning predictive pattern model with at least a subset of the determined signatures as inputs and associated class labels as outputs comprises at least one of:
(i) splitting the determined signatures into a training sub-dataset and a testing sub-dataset, wherein the training sub-dataset is the subset of the determined signatures;
(ii) training the machine learning predictive pattern model with a K-Nearest Neighbor (KNN) and Support Vector Machine (SVM) supervised-learning methodology by feeding determined signatures from the training sub-dataset into the at least one machine learning predictive pattern model as inputs and associated class labels as outputs;
(iii) validating the trained machine learning predictive pattern model with the testing sub-dataset; and
(iv) fine-tuning KNN and SVM model parameters.

### Claim 11 (depends on 1)

The method of claim 1 further comprising:
automatically receiving sensor data from an ongoing batch production run of the industrial process; and
processing the received sensor data with the trained machine learning predictive pattern model to determine a prediction of output of the ongoing batch production run as conforming or non-conforming with the operational standards.

### Claim 12 (depends on 11)

The method of claim 11 wherein the determined prediction includes an indication of statistical probability in the determined prediction.

### Claim 13 (depends on 12)

The method of claim 12 further comprising:
issuing an alert to a user if both (i) the determined prediction indicates the output is non-conforming and (ii) the indication of statistical probability in the determined prediction is above a threshold.

### Claim 14 (depends on 12)

The method of claim 12 further comprising:
at a current batch progress stage of the ongoing batch production run, examining the indication of statistical probability in the determined prediction; and
performing an online root-cause analysis if: (i) the determined prediction indicates the output is non-conforming and (ii) the indication of statistical probability in the determined prediction is above a threshold.

### Claim 15 (depends on 12)

The method of claim 12 wherein (i) the determined prediction indicates the output is non-conforming and (ii) the indication of statistical probability in the determined prediction is above a threshold, and the method further comprises:
from among the historical operating data, determining at least one K-nearest neighbor batch to the ongoing batch production run; and
based on the determined at least one K-nearest neighbor batch, performing a comparative analysis between the at least one K-nearest neighbor batch and a standard reference batch by using at least one multivariate statistical model.

### Claim 16 (depends on 1)

The method of claim 1 further comprising:
deploying the trained machine learning predictive pattern model online in the industrial process at one or more pre-specified time points of a batch run;
using the deployed trained machine learning predictive pattern model, predicting whether output of the batch run of the industrial process will conform or not conform with the operational standards; and
executing real-time batch monitoring and analysis based on the predicting, wherein the analysis includes (i) diagnosing one or more operational problems in the batch run of the industrial process and (ii) providing prescriptive guidance to a plant operator with one or more recommended corrective actions.

### Claim 17 (depends on 16)

The method of claim 16 wherein diagnosing one or more operational problems and providing prescriptive guidance to a plant operator further comprises:
identifying one or more contributing KPIs; and
outputting an alert to the plant operator with an associated risk assessment report and a root-cause analysis report.

### Claim 18 (depends on 1)

The method of claim 1 wherein the operational standard is at least one of: a physical criterion and a chemical criterion.

### Claim 19

A computer system for creating a machine learning predictive model for real-world batch production industrial process monitoring and optimization, the system comprising:
a processor; and
a memory with computer code instructions stored thereon, the processor and the memory, with the computer code instructions, being configured to cause the system to:
obtain historical operating data from a plurality of batch production runs of an industrial process;
standardize the obtained historical operating data for each run of the plurality of batch production runs;
for each batch production run of the plurality: (i) partition standardized operating data corresponding to the batch production run into one or more stages and (ii) determine one or more signature for each of the one or more stages using the partitioned standardized operating data corresponding to the one or more stages;
associate each determined signature with a class label based upon output of a batch production run of the plurality corresponding to the determined signature conforming with operational standards or not conforming with the operational standards; and
train a machine learning predictive pattern model with at least a subset of the determined signatures as inputs and associated class labels as outputs, wherein the training configures the model to predict, based on operating data from a real-world batch production process, whether output of the real-world batch production process will conform or not conform with the operational standards.

### Claim 20

A computer program product for creating a machine learning predictive model for real-world batch production industrial process monitoring and optimization, the computer program product comprising:
one or more non-transitory computer-readable storage devices and program instructions stored on at least one of the one or more storage devices, the program instructions, when loaded and executed by a processor, cause an apparatus associated with the processor to:
obtain historical operating data from a plurality of batch production runs of an industrial process;
standardize the obtained historical operating data for each run of the plurality of batch production runs;
for each batch production run of the plurality: (i) partition standardized operating data corresponding to the batch production run into one or more stages and (ii) determine one or more signature for each of the one or more stages using the partitioned standardized operating data corresponding to the one or more stages;
associate each determined signature with a class label based upon output of a batch production run of the plurality corresponding to the determined signature conforming with operational standards or not conforming with the operational standards; and
train a machine learning predictive pattern model with at least a subset of the determined signatures as inputs and associated class labels as outputs, wherein the training configures the model to predict, based on operating data from a real-world batch production process, whether output of the real-world batch production process will conform or not conform with the operational standards.
