Detection of harmful process intentions in intention-based production processes
By introducing machine-readable models and security libraries into the production process, combined with expert review, the vulnerability of intent-based production processes to attacks has been addressed. This enables the detection and protection against malicious intent, thereby improving the security and stability of the production process.
Patent Information
- Application Number
- CN202010479641.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Priority Date
- 2019-05-31
- Filing Date
- 2020-05-29
- Publication Date
- 2024-11-08
- Estimated Expiration
- 2040-05-29
AI Technical Summary
Existing intent-based production processes are vulnerable to being misled by attackers, leading to threats to functional security and productivity, and there is a lack of effective means to detect malicious intent.
By providing machine-readable production models, linking intent models and process models, and combining security libraries and expert reviews, we assess potentially harmful process intents, provide reports to review intent changes, and ensure that safe process intents are maintained outside of conventional engineering mechanisms.
It enables effective detection and protection against harmful intentions, improves the safety and stability of the production process, and reduces the risks caused by misleading intentions.
Smart Images

Figure CN112015148B_ABST
Abstract
Description
Technical Field
[0001] The invention relates to a method, a control device, a computer program and a computer-readable storage medium for the detection of harmful process intentions in an intention-based production process. Background Art
[0002] Automated systems are moving towards increasing levels of autonomy, where operational target KPIs or product design specifications automatically reconfigure the automated system based on codified knowledge about the production process. This intent-based approach represents a large quality improvement and huge efficiency lever, but the same efficiency can be misled intentionally by an attacker or accidentally by mistake; new attack surfaces are opening up regarding functional safety and productivity.
[0003] It is therefore an object of the present invention to provide an improved intent-based production process, wherein preferably harmful intent can be detected. Summary of the invention
[0004] According to one aspect, a method for detecting harmful process intent in an intent-based production process includes the following steps: providing a machine-readable production model, which links an intent model and a process model, the intent model is related to the process intent, the process intent includes production process functions, constraints on measurable attributes of production process functions and / or a sequence of production process functions required for the production process, and the process model is related to process knowledge including production process behavior; determining at least one security library of safety process intent, wherein the safety library is protected by a safety mechanism for maintenance outside of conventional intent-based engineering mechanisms and operating mechanisms; evaluating changes in process intent based on at least one safety process intent, taking into account potential harmful process intent; and providing a report of potential harmful process intent based on the potential harmful process intent for review.
[0005] The term "potentially harmful process intent and / or harmful process intent" as used herein relates to process intent that causes harm in a production process. In other words, harmful process intent relates to production process functions, restrictions on measurable attributes of production process functions, and / or sequences of production process functions required for a production process, which may result in harm to productivity, such as damage and / or destruction of production process products / materials, and / or harm to functional safety, such as damage and / or destruction of production process hardware (such as production process modules of production process devices that control materials, such as pumps, mixers, tanks, reactors, heat exchangers, etc.), or harm to human health or life. Therefore, harmful process intent covers process intent that is created maliciously or mistakenly.
[0006] The term "safety process intent" as used herein relates to process intent related to production process safety. In other words, the safety process intent relates to production process functions, constraints on measurable attributes of production process functions and / or sequences of production process functions required for the production process, which are known to endanger productivity, such as damage and / or destruction of production process products / materials, and / or endanger functional safety, such as damage and / or destruction of production process hardware (such as production process modules of production process devices that control materials, such as pumps, mixers, tanks, reactors, heat exchangers, etc.), or are harmful to human health or life. In other words, the safety process intent plays the role of a safety intention rule.
[0007] Preferably, evaluating the change of the process intent includes checking whether the changed process intent has a potentially harmful process intent, thereby evaluating the potentially harmful process intent according to the safety process intent in the safety library.
[0008] A change in process intent preferably involves reconfiguring and / or changing the engineering baseline to be deployed for the production process.
[0009] Preferably, the safety process is intended for safety protection and / or productivity protection.
[0010] Preferably, the security library is provided externally.
[0011] Preferably, the production model is validated, preferably depending on key performance indicators of the production process, in particular the product produced, to determine validation data. Based on the validation data, the process intent may be changed. The change in the process intent is then checked for potentially harmful process intent.
[0012] During validation, the production process is preferably tested for any violations of the intended safety process.
[0013] Therefore, core safety functions are separated out of the scope of intent-based engineering.
[0014] Preferably, a dedicated set of security intent rules is defined, in particular security intent rules used during verification. These security intent rules are not available in conventional engineering processes.
[0015] Thus, an improved intent-based production process is provided, wherein preferably harmful intent can be detected.
[0016] In a preferred embodiment, the potentially harmful process intent includes: a change in which a new process intent contradicts an existing process intent, and / or a change in which a new process intent contradicts a process intent removed in a previous change.
[0017] Thus, an improved intent-based production process is provided, wherein harmful intent can be preferentially detected.
[0018] In a preferred embodiment, the method includes the step of manual review of the report of potentially harmful process intent by a process expert.
[0019] Preferably, the report is reviewed using the four-eyes principle, in particular by two process experts.
[0020] Thus, an improved intent-based production process is provided, wherein harmful intent can be preferentially detected.
[0021] Preferably, review of the report includes acceptance or rejection of the potentially harmful process intent.
[0022] Thus, an improved intent-based production process is provided, wherein harmful intent can be preferentially detected.
[0023] In a preferred embodiment, the security library includes a general security library and a specific project security library.
[0024] For example, the library contains hazardous states of chemical reactions, substances that cannot be mixed, environments that should always be observed and / or substances that cause explosions or overheating.
[0025] Thus, an improved intent-based production process is provided, wherein harmful intent can be preferentially detected.
[0026] In a preferred embodiment, the method comprises the following steps: automatically determining at least one safety process intention from a general safety library applicable to the production process according to the current production process project as a specific project safety process intention for the specific project safety library.
[0027] Preferably, from the production process project, in particular the production process project file, the different steps covered by the production process, such as the chemical reactions to be performed, can be derived manually and / or automatically.
[0028] Thus, an improved intent-based production process is provided, wherein harmful intent can be preferentially detected.
[0029] In a preferred embodiment, the method comprises the step of classifying a specific project process intention into one of the categories "proven good intention", "proven negative impact" and / or "proven negative process behavior".
[0030] For example, a process classified as “proven adverse impact” is intended to include a description that any temperature above 100°C will cause the production process to stop.
[0031] Thus, an improved intent-based production process is provided, wherein harmful intent can be preferentially detected.
[0032] In a preferred embodiment, the method comprises the step of cryptographically securing the assessment of the intended change of the process.
[0033] Thus, an improved intent-based production process is provided, wherein harmful intent can be preferentially detected.
[0034] In a preferred embodiment, the method comprises the step of continuously extending the safety library based on the project-specific process intent of the checked changes.
[0035] Thus, an improved intent-based production process is provided, wherein harmful intent can be preferentially detected.
[0036] In a preferred embodiment, the method comprises the step of classifying the potentially harmful intent into different levels and / or categories of harm, preferably for illustration purposes in a report of the potentially harmful process intent.
[0037] Thus, an improved intent-based production process is provided, wherein harmful intent can be preferentially detected.
[0038] In a preferred embodiment, the method comprises the following steps: Providing a generic library of negative production intentions.
[0039] Preferably, negative production intentions include hazardous conditions, in particular for chemical reactions, in other words substances that cannot be mixed, environments that should always be observed and / or substances that cause explosions or overheating.
[0040] Thus, an improved intent-based production process is provided, wherein harmful intent can be preferentially detected.
[0041] In a preferred embodiment, the method comprises the step of checking for potential harmful intentions based on virtual testing using a simulation model of the process.
[0042] Malicious intent is therefore detected as part of an automated intent verification process embedded into the engineering workflow. As a result, security process intent becomes a constant part of the verification process.
[0043] Furthermore, virtual testing is used to check whether the simulated production process exhibits any behavior that violates the intent of the safety process.
[0044] Thus, an improved intent-based production process is provided, wherein harmful intent can be preferentially detected.
[0045] According to another aspect, a control device is provided, the control device being configured to perform the method described herein.
[0046] According to another aspect, a computer program is provided, which, when executed on a control device as described herein, instructs the control device to perform the steps of the method as described herein.
[0047] According to another aspect, a computer readable storage medium is provided, the storage medium being configured to store the computer program described herein. BRIEF DESCRIPTION OF THE DRAWINGS
[0048] The subject matter of the invention will be explained in more detail hereinafter with reference to preferred exemplary embodiments, which are illustrated in the accompanying drawings, in which:
[0049] Figure 1 A control device is schematically shown; and
[0050] Figure 2 A method for detection of deleterious process intent in an intent-based production process is presented. DETAILED DESCRIPTION
[0051] The reference symbols used in the drawings and their meanings are listed in summary form in the list of reference symbols. In principle, identical components are denoted by the same reference symbols in the figures.
[0052] Figure 1 A system including an apparatus 10, a machine learning unit 20, a simulation unit 30 and a production process module 40 is shown.
[0053] The control device 10 includes a model unit 11, a control unit 12 and a safety unit 13. The device 10 is provided with an intention model I and a process model P. The intention model I is associated with a process intention Ip, which includes production process functions, constraints on measurable attributes of production process functions and / or a sequence of production process functions required for the production process. The process model P is associated with process knowledge including production process behavior. The model unit 11 determines a production model M that links the intention model I to the process model P.
[0054] Therefore, as the process intent Ip is integrated into the production model M, so-called intent-based engineering allows designers and operators to focus on "what" the process should do, in other words, the goal, rather than "how" it should be achieved, in other words, not the means of implementation. By hiding the complexity of automation technology, intent-based automation enables non-automation experts to effectively automate and optimize their processes from an operational perspective. In addition, formalized goals can be subject to automation verification, and they help to improve the coverage of automation.
[0055] The production model M is provided to the control unit 12 which determines a control strategy S for controlling the production process module 40. The production process module 40 itself is a physical module which is provided with an input material 41 and processes the input material 41 into an output material 42. During the processing some measurement data 43 may be obtained from the production process module 40, such as the temperature.
[0056] The measurement data 43 are provided to the machine learning unit 20, which verifies and / or validates the current production model M, thereby generating online verification data Von. These online verification data Von are provided back to the model unit 11, which then adjusts the current production model M based on the online verification data Von. Therefore, the production model M can be iteratively improved.
[0057] Furthermore, the production model M is provided to the simulation unit 30. The simulation unit comprises a digital twin of the production process module 40 and verifies and / or validates the current production model M, thereby generating offline validation data Voff. The offline validation data Voff is also provided back to the model unit 11, which then adjusts the current production model M based on the offline validation data Voff. Therefore, the production model M can be iteratively improved.
[0058] In the illustrative example, production process module 40 is a fermenter.
[0059] The process module 40 should be able to maintain a specific temperature and protect the fermentation process from overheating. This is achieved by describing cause and effect, such as opening a heating circuit when the temperature is too low, or opening an emergency valve when the temperature is too high. Encapsulating this basic equipment control and protection is a core feature of the process module 40.
[0060] For a hypothetical fermentation process, the ideal temperature is 50°C, which is to be maintained for a duration of 2 days. A corresponding production sequence can be generated to maintain this process step for two days, parameterizing the mixing and heating services of the tank modules to maintain 50°C.
[0061] To obtain an acceptable fermented product quality, the process must not exceed 70°C or fall below 40°C. In the first case, the yeast used will die and the batch must be discarded. In the second case, the fungi will become inactive and the fermentation will stop. However, unlike overheating, if the reaction has not stopped for more than 2 hours (because then the product has already started to deteriorate), the process can be continued by raising the temperature again. The module is configured to protect the process according to these limits, overriding the given default values.
[0062] The fermentation speed also changes with a specific temperature within the permissible temperature range. For every +1°C, 1 hour of production time can be gained. With this knowledge, production can be scaled up without compromising product quality.
[0063] Finally, the relevant information is captured from the process topology. The fermenter requires an external heat source, which is provided by the second module, the heat exchanger. However, the actual intention is not to achieve a specific temperature inside the heat exchanger (which can be measured directly by the heat exchanger), but to achieve a specific temperature inside the fermenter, which knows its temperature without directly controlling the heat exchange process. This leads to instabilities in the temperature control. Therefore, the rate at which the temperature decreases with the known fermentation time is not sufficient to maximize the production volume, because at the edge of the good operating area, the control instability will cause the temperature to rise above the maximum allowed level and the batch will be lost. In order to enable the process to get closer to its boundaries, the temperature control must be stabilized, for example by having the heat exchanger use the temperature measurements from the tank directly, but first, the impact of process connectivity on control stability must be explicitly captured.
[0064] The traditional or modular engineering approach specifies only a set point of 50°C and a corresponding warning limit. By combining additional process knowledge in the form of process models P, which can be easily captured by process designers on one or more modules, a production model M can be constructed, which is associated with process intent (such as production KPIs) in the form of intent models I.
[0065] In the absence of such optimization, the operating range of the process must be determined through a series of trials and errors. This is usually done collaboratively by operators and process experts during the early production phase after initial plant startup, as no other codified information about the process behavior is available.
[0066] Based on the production model M, the automation system knows out of the box that even without stable control, production time can be reduced by 30% without any risk to product quality; by optimizing the stability of temperature control, production time can even be safely reduced to 50%. Secondly, the operating strategy S can be motivated by tracing the automation decisions back to the process knowledge provided by the process experts in the same company.
[0067] Whenever the process intention Ip is changed by changing the provided intention model I or based on the online verification data Von and / or the offline verification data Voff, the changed process intention Ip is provided to the safety unit 13. The safety unit 13 includes an evaluation unit 13a and a reporting unit 13b. The evaluation unit 13a checks whether the provided changed process intention Ip has a potentially harmful intention Ih. Therefore, the evaluation unit 13a evaluates the changed process intention Ip according to the safety process intention Is, which is provided to the safety unit 13, in particular the evaluation unit 13a, by the safety library L. The safety process intention Is involves the production process function, the constraints on the measurable attributes of the production process function and / or the production process function sequence required by the production process, which are known to endanger productivity, such as damage and / or destruction of the production process product / material, and / or endanger functional safety, such as damage and / or destruction of the hardware of the production process (such as the production process module of the production process device that controls the material, such as pumps, mixers, tanks, reactors, heat exchangers, etc.), or is harmful to human health or life.
[0068] The determined potentially harmful intention Ih is then provided by the evaluation unit 13a to the reporting unit 13b. The reporting unit 13b determines a report R for audit, thereby presenting the determined potentially harmful intention Ih. The report R is provided to the technical expert 50, in particular a safety expert for audit. Therefore, safety-critical changes to the process intention Ip are not automatically approved, but must be approved manually. The technical expert 50 then reads the report R and decides whether the potentially harmful intention Ih is indeed harmful to the process or whether it can be approved. If the technical expert 50 approves the change to the process intention Ip, the intention model I associated with the process intention Ip is adjusted accordingly. Otherwise, the change to the process intention Ip will be canceled. Information related to the decision of the technical expert 50 is included in the evaluation data E, which is provided to the model unit 11 by the technical expert 50, in particular via the human-machine interface of the control device 10.
[0069] Figure 2 A flow chart of a method for detection of deleterious process intent in an intent-based production process is shown.
[0070] In step S10, a machine-readable production model M is provided, which links the intention model I and the process model P. The intention model I is related to the process intention Ip, and the process intention Ip includes production process functions, constraints on measurable attributes of production process functions and / or the sequence of production process functions required for the production process. The process model P is related to process knowledge including production process behavior.
[0071] In step S20 , a safety repository L of at least one safety process intent Is is determined, wherein the safety repository L is protected by a safety mechanism to be maintained outside of conventional intent-based engineering mechanisms and operating mechanisms.
[0072] In step S30, changes in the process intention Ip are evaluated based on at least one safe process intention Is, taking into account potentially harmful process intentions Ih.
[0073] In step S40, a report of the potentially harmful process intention Ih is provided based on the potentially harmful process intention Ih for review.
[0074] Reference numerals:
[0075] 10 Control Equipment
[0076] 11 Model Unit
[0077] 12 Control Unit
[0078] 13 Security Unit
[0079] 13a Evaluation Unit
[0080] 13b Reporting Unit
[0081] 20 Machine Learning Units
[0082] 30 analog units
[0083] 40 Production process modules
[0084] 41 Input Materials
[0085] 42 Output Materials
[0086] 43 Measurement data
[0087] Von Online Verification Data
[0088] Voff Offline Verification Data
[0089] I Intent Model
[0090] P Process Model
[0091] M Production Model
[0092] S Control Strategy
[0093] Ip process intention
[0094] Is Security Process Intent
[0095] Ih Potentially Harmful Intent
[0096] R Report
[0097] E. Evaluation Data
[0098] S10 provides machine-readable production models
[0099] S20 Confirmed Security Library
[0100] S30 Evaluate changes in production intentions
[0101] S40 Provides reporting of potentially harmful process intent
Claims
1. A method for detecting harmful process intent in an intent-based production process, comprising the following steps: Providing (S10) a machine-readable production model (M), the machine-readable production model (M) links an intention model (I) and a process model (P), the intention model (I) is associated with a process intention (Ip), the process intention (Ip) includes production process functions, constraints on measurable attributes of the production process functions and / or a sequence of production process functions required for the production process, and the process model (P) is associated with process knowledge including production process behavior; determining (S20) a safety library (L) of at least one safety process intent (Is), wherein the at least one safety process intent (Is) relates to a process intent (Ip) related to the safety of the production process, wherein the safety library (L) is protected by a safety mechanism to be maintained outside of conventional intent-based engineering mechanisms and operating mechanisms; validating the production model (M), depending on key performance indicators of the production process of the product being produced, thereby determining validation data, based on which the process is intended to be changed; evaluating (S30) a change of a process intention (Ip) taking into account a potentially harmful process intention (Ih) according to the at least one safe process intention (Is); as well as A report of the potentially harmful process intention (Ih) is provided (S40) for review based on the potentially harmful process intention (Ih).
2. A method according to any one of the preceding claims, wherein Potentially harmful process intentions (Ih) include: Changes where the new process intent contradicts the existing process intent, and / or changes where the new process intent contradicts the process intent removed in a previous change.
3. The method according to any one of the preceding claims, comprising the steps of: The report of potentially harmful process intentions (Ih) is manually reviewed by process experts.
4. The method according to claim 1 or 2, wherein The security library (L) includes a general security library and a specific project security library.
5. The method according to claim 4, comprising the steps of: According to the current production process project, at least one safety process intention (Is) is automatically determined from the general safety library applicable to the production process as a specific project safety process intention for the specific project safety library.
6. The method according to claim 5, comprising the steps of: The specific project safety process intent is classified into one of the categories of "proven good intentions", "proven negative impacts" and / or "proven negative process behaviors".
7. The method according to any one of claims 1, 2, 5 and 6, comprising the following steps: The evaluation of the change of the process intent (Ip) is cryptographically secured.
8. The method according to any one of claims 1, 2, 5 and 6, comprising the steps of: The safety library (L) is continuously expanded based on the project-specific safety process intentions of the examined changes.
9. The method according to any one of claims 1, 2, 5 and 6, comprising the following steps: The potentially harmful intentions (Ih) are classified into different hazard levels and / or categories for descriptive purposes in the report of potentially harmful process intentions.
10. The method according to any one of claims 1, 2, 5 and 6, comprising the steps of: Provides a general library of negative production intents.
11. The method according to any one of claims 1, 2, 5 and 6, comprising the steps of: The potentially harmful intentions (Ih) are checked based on virtual testing using a simulation model of the process.
12. A control device configured to execute the method according to any one of claims 1 to 11. 13 . A computer program product, when the computer program product is executed on the control device according to claim 12 , instructs the control device to execute the steps in the method according to any one of claims 1 to 11 .
14. A computer-readable storage medium configured to store the computer program product according to claim 13.
