Method and system for monitoring a process
The invention provides a way to monitor a process ( 1 ). The invention utilizes explanation values. A ML model of the process is created. The ML model utilizes measurements from the process as inputs to the ML model and forms model outputs. The inputs are classified to groups in the. The explanation values are calculated for each input indicating contribution of the input to the model output/s, and a sum of the explanation values for each group are calculated. The calculated sums are monitored, each sum indicating status of the group in question.
1. A method for monitoring a process and/or identifying source(s) of process disturbances or quality defects comprising:
creating a machine learning (ML) model of the process, which ML model utilizes measurements from the process as inputs to the ML model and forms model outputs, and classifying the inputs to groups,
calculating an explanation value of machine learning for each input indicating contribution of the input to the model output/s, and calculating a sum of indication values for each group, the indication values relating to the explanation value of machine learning,
monitoring the calculated sums, each sum indicating status of the group in question, and
controlling, optimizing and/or troubleshooting the process based on the status of one of more of the groups.
2. The method according to claim 1 , wherein classifying the inputs to groups utilizes a predetermined database containing info about how the inputs belong to the groups, and monitoring the calculated sums is made in a group specific way.
3. The method according to claim 1 wherein the explanation value of machine learning is a Shapley additive explanations (SHAP) value, a value from a local interpretable model-agnostic explanations (LIME) method, a value from a Deep Learning Important FeaTures (deepLIFT) method.
4. The method according to claim 3 , wherein the groups comprise one or more of main groups and optionally one or more of subgroups.
5. The method according to claim 1 wherein the indication values are explanation values of machine learning or deviations of the explanation values, the deviations of the explanation values being differences between the explanation values and normal explanation values.
6. The method according to claim 5 , wherein the groups are pre-defined.
7. The method according to claim 6 , wherein the status of the group is defined utilizing at least one threshold value for said sums.
8. The method according to claim 7 , wherein the status of the group is fine when the sum is negative or relatively near to zero, and the status of the group is not fine when the sum is positive.
9. The method according to claim 1 , wherein the process is a pulp process, papermaking process, board making process, tissue making process, paper machine, pulp mill, tissue machine, board machine, water treatment process, waste water treatment process, raw water treatment process, water re-use process, any industrial water treatment process, municipal water, municipal waste water treatment process, sludge treatment process, mining process, or oil recovery process-or any other industrial process.
10. The method according to claim 4 , wherein the process is a paper, board or tissue making process, and the main groups comprise one or more of production, chemistry, or towers.
11. The method according to claim 10 , wherein the subgroups comprise one or more of quality of incoming pulp(s), anionic trash in incoming pulp(s), hydrophobic contaminants in incoming pulp(s), microbial activity in incoming pulp(s), wet end chemistry, microbial activity in wet end, anionic trash (amount or concentration of dissolved anionic compounds) in wet end, hydrophobic contaminants in wet end, retention, sizing, the quality of broke stream(s), hydrophobic contaminants in broke stream(s), anionic trash in broke stream(s), microbial activity in broke stream(s), the quality of raw water, the quality of water circulation, grade, grammage, the amount of produced product, breaks/shutdowns of the process.
12. The method according to claim 1 , wherein the method comprises at least one next step for providing an action proposal for corrective actions based on the status of one of more of the groups.
13. The method according to claim 1 , wherein the controlling and/or optimizing step comprises one or more of controlling dosing amount of chemicals, dosing points of chemicals, dosing intervals of chemicals, selection of chemical types to be used in the process, process conditions, such as pH, temperature, flow rate of process streams, and process stream delays, such as pulp, broke or water stream delays in process equipment, such as in towers, tanks, pulpers, basins or process equipment.
14. The method according to claim 1 , wherein the inputs comprise real time values.
15. A system to monitor a process comprising a machine learning (ML) module in order to create a ML model of the process, which ML model utilizes measurements from the process as inputs to the ML model and forms model outputs, wherein the system further comprises a classification module in order to classify the inputs to groups,
a calculation module in order to calculate explanation value of machine learning for each input indicating contribution of the input to the model output/s, and calculating a sum of indication values for each group, the indication values relating to the explanation values of machine learning and
a monitoring module to monitor the calculated sums, each sum indicating status of the group in question, and
wherein the system is configured to control, optimize and/or troubleshoot the process based on the status of one of more of the groups.
16. The system according to claim 15 , wherein the classification module in order to classify the inputs to groups utilizes a predetermined database containing info about how the inputs belong to the groups, and the monitoring module to monitor the calculated sums is arranged to made in a group specific way.
17. The system according to claim 16 , wherein the explanation value of machine learning is a Shapley additive explanations (SHAP) value, a value from a local interpretable model-agnostic Explanations (LIME) method, a value from a Deep Learning Important FeaTures (deepLIFT) method.
18. The system according to claim 17 , wherein the groups comprise main groups and subgroups.
19. The system according to claim 17 wherein the indication values are the explanation values or deviations of the explanation values, the deviations of the explanation values being differences between the explanation values and normal explanation values.
20. The system according to claim 19 , wherein the system comprises a library module to keep the normal explanation values.
21. The system according to claim 19 , wherein the groups are pre-defined.
22. The system according to claim 21 , wherein the status of the group is fine when the sum is negative or relatively near to zero, and the status of the group is not fine when the sum is positive.
23. The according to claim 21 , wherein the process to be monitored is a paper, board or tissue making process, and the main groups comprise one or more of production, chemistry, or towers.
24. The system according to claim 23 , wherein the subgroups comprises one or more of quality of incoming pulp(s), anionic trash in incoming pulp(s), hydrophobic contaminants in incoming pulp(s), wet end chemistry, microbial activity in wet end, anionic trash (amount or concentration of dissolved anionic compounds) in wet end, hydrophobic contaminants in wet end, retention, sizing, the quality of broke stream, hydrophobic contaminants in broke stream, anionic trash in broke stream, the quality of raw water, the quality of water circulation, grade, grammage, the amount of produced product, breaks/shutdowns of the process.
25. The system according to claim 15 , wherein in that the inputs are real time values.