EP-4741966-A2 - CHEMICAL PRODUCTION

EP4741966A2EP 4741966 A2EP4741966 A2EP 4741966A2EP-4741966-A2

Abstract

The present teachings relate to a method for improving a production process for manufacturing a chemical product using at least one input material at an industrial plant, the industrial plant comprising a plurality of physically separated equipment zones, the method comprising: providing, via an interface, an upstream object identifier comprising input material data; receiving, at a computing unit, real-time process data from one or more of the equipment zones; determining, via the computing unit, a subset of the real-time process data based on the upstream object identifier and a zone presence signal; computing, via the computing unit, at least one zone-specific performance parameter of the chemical product related to the upstream object identifier based on the subset of the real-time process data and historical data; determining, in response to at least one of the performance parameters, a target equipment zone where the input material and/or chemical product is to be sent. The present teachings also relate to a system, and a software program.

Inventors

WINKLER, Christian-Andreas
RUDOLPH, HANS
HARTMANN, MICHAEL
RAUTENSTRAUCH, Markus
HUANG, Yuan En
WANDERNOTH, Sebastian
YAKUT, Nataliya

Assignees

BASF SE

Dates

Publication Date: 20260513
Application Date: 20211210

Claims (20)

A method for improving a production process for manufacturing a chemical product at an industrial plant, the industrial plant comprising a plurality of physically separated equipment zones and one or more computing units, and the product being manufactured by processing, via the plurality equipment zones, at least one input material using the production process, which method comprises: - providing, via an interface, an upstream object identifier comprising input material data; wherein the input material data is indicative of one or more properties of the input material, - receiving, at any of the computing units, real-time process data from one or more of the equipment zones; wherein the real-time process data comprises real-time process parameters and/or equipment operating conditions, - determining, via any of the computing units, a subset of the real-time process data based on the upstream object identifier and a zone presence signal; wherein the zone presence signal indicates presence of the input material at a specific equipment zone during the production process, - computing, via any of the computing units, at least one zone-specific performance parameter of the chemical product related to the upstream object identifier based on the subset of the real-time process data and historical data, wherein the computing of the at least one zone-specific performance parameter is done using a model, which is at least partly an analytical computer model and /or at least partly one or more machine learning ("ML") models; - determining, in response to at least one of the performance parameters, a target equipment zone where the input material and/or chemical product is to be sent.
The method of claim 1, wherein the input material for the processing via the equipment is divided into at least two packages wherein the size of a package is fixed or is determined based on an input material weight or amount, for which considerably constant process parameters or equipment operation parameters can be provided by the equipment.
The method of any one of the preceding claims, wherein the processing of the at least two packages is managed by means of corresponding data objects, each of which at least including an historical object identifier.
The method of any one of the preceding claims, wherein the method also comprises: - appending, to the upstream object identifier, the at least one zone-specific performance parameter.
The method of any one of the preceding claims, wherein the historical data comprise data from one or more historical object identifiers related to previously processed input material.
The method of claim 5, wherein at least of the one historical object identifiers is appended with at least a part of that historical process data which is indicative of the process parameters and/or equipment operating conditions that the previously processed input material was processed under.
The method of any one of the preceding claims, wherein the machine learning model is trained using the historical data.
The method of any one of the preceding claims, wherein the machine learning model is trained using the historical data from the one or more historical object identifiers.
The method of any one of the preceding claims, wherein at least one of the zone-specific performance parameter is also related to a derivative material that is produced from the input material, but prior to the chemical product during the production process.
The method of any one of the preceding claims, wherein the machine learning model is a data driven model, in particular a regression model, or wherein the machine learning model is a hybrid model.
The method of any one of the preceding claims, wherein the industrial plant comprises an Internet-of-Things (IoT) Edge device or component and wherein the underlying ML system is implemented to find or create an algorithm, which is deployed to the loT Edge device or component, in order to use the accordingly created or found algorithm for controlling the loT Edge device.
The method of any one of the preceding claims, further comprising: providing an abstraction layer which includes an object database and which serves as an abstraction layer for the production equipment, for the corresponding input materials and for package-related data.
The method of the preceding claim, wherein the abstraction layer connects to certain processing and/or ML components within a Cloud computing platform, wherein for this connection, a data streaming protocol is used, and wherein streamed and received product data is used by the ML system to find or create algorithms for getting additional data related to an underlying chemical product.
The method of the preceding claim, wherein the additional data concern predictable product quality control (QC) data of the underlying chemical product.
The method of any one of the preceding claims, wherein the training data for training the ML model also comprise historical and/or current laboratory test data, or data from the past and/or recent samples, said historical and/or current laboratory test data being indicative of the performance parameters of the chemical product.
The method of any one of the preceding claims, wherein the model or the ML model provides at least one confidence value indicative of the confidence level for the at least one zone-specific performance parameter.
The method of the preceding claim, wherein the confidence is appended to the upstream object identifier, in particular as metadata.
The method of one of the two preceding claims, wherein, if the confidence level of the prediction or computation of at least one zone-specific performance parameter fall below an accuracy threshold value, a warning is triggered at the control system for production.
The method of any one of the three preceding claims, wherein, in response to the confidence level of the prediction or computation of at least one zone-specific performance parameter falling below the accuracy threshold value, a sampling object identifier is automatically provided, e.g, via the interface.
A computer program, or a non-transitory computer readable medium storing the program, comprising instructions which, when executed by suitable one or more computing units, cause the computing units to carry out the method steps of any of the above method claims.

Description

TECHNICAL FIELD The present teachings relate generally to computer assisted chemical production. BACKGROUND ART In industrial plants, input material is processed to manufacture one or more products. Properties of the manufactured product thus have a dependence upon manufacturing parameters. It is usually desired to correlate manufacturing parameters to at least some properties of the product for ensuring product quality or production stability. Within process industry, or industrial plants such as chemical or biological production plants, one or more input materials are processed using a production process for producing one or more chemical or biological products. Production environment in the process industry can be complex, accordingly the properties of the product may vary according to variations in the production parameters that influence said properties. Usually the dependency of the properties on production parameters can be complex and intertwined with a further dependence on one or more combinations of different parameters. In some cases, the production process may be divided into multiple stages, which can further aggravate the problem. It may thus be challenging to produce a chemical or biological product with consistent and/or predictable quality. For keeping quality of the chemical product consistent, quality control may be performed. Quality control normally involves collecting one or more samples of the chemical product after or during the production process. The samples are then analyzed and then corrective action, as necessary, may be taken. To be effective, the samples may need to be collected regularly and should be representative of the statistical variation of the chemical product. Dependent upon the frequency of variations occurring in the production process, the frequency of quality control may be required to be aligned. Quality control can thus be expensive and time consuming. Production processes with narrower variations can require expensive equipment and/or input material. Further, in contrast to discrete processing, chemical or biological processing such as continuous, campaign or batch processes, may provide vast amounts of time series data. However, machine learning via traditional time series approaches has proven to be less practical, since it can be difficult to integrate data according to the needs of horizontal integration across the value chain. In particular, easy and meaningful data exchange or standardization can pose major problems. There is hence a need for approaches that can improve quality and production stability across the value chain ideally from barrel to end product. SUMMARY At least some of the problems inherent to the prior art will be shown solved by the subject matter of the accompanying independent claims. At least some of the further advantageous alternatives will be outlined in the dependent claims. When viewed from a first perspective, there can be provided a method for improving a production process for manufacturing a chemical product at an industrial plant, the industrial plant comprising a plurality of physically separated equipment zones and one or more computing units, and the product being manufactured by processing, via the plurality equipment zones, at least one input material using the production process, which method comprises: providing, via an interface, an upstream object identifier comprising input material data; wherein the input material data is indicative of one or more properties of the input material,receiving, at any of the computing units, real-time process data from one or more of the equipment zones; wherein the real-time process data comprises real-time process parameters and/or equipment operating conditions,determining, via any of the computing units, a subset of the real-time process data based on the upstream object identifier and a zone presence signal; wherein the zone presence signal indicates presence of the input material at a specific equipment zone during the production process,computing, via any of the computing units, at least one zone-specific performance parameter of the chemical product related to the upstream object identifier based on the subset of the real-time process data and historical data;determining, in response to at least one of the performance parameters, a target equipment zone where the input material and/or chemical product is to be sent. Hence, the input material and/or chemical product can be diverted or destined to the target equipment zone. The target equipment zone may be one of a plurality of target equipment zones. The determination may be made by any one or more of the computing units. The computing unit may even be configured to divert the material and/or the product to the target zone, for example by controlling one or more actuators. The method steps may be performed by the same computing unit or different computing units operatively coupled to each other, for example, if one of the uni