Analysis method and device for the same

Abstract

According to the present invention, in order to provide a method for analyzing defects in an industrial process plant, e.g., a painting plant, defect situations can be analyzed simply and reliably by the present method, and the present method comprises: automatic recognition of defect situations, particularly in an industrial process plant (101); storage of a defect situation data set for each recognized defect situation in a defect database (136); automatic determination of the cause of the defect for each recognized defect situation and / or automatic determination of process values ​​related to the defect situation based on the defect data set for each recognized defect situation.

Description

Technology Field

[0001] The present invention relates to a method for analyzing defects in an industrial process plant, for example, a painting plant. means of solving the problem

[0002] The object of the present invention is to provide a method for analyzing defects in an industrial process plant, for example, a painting plant, wherein the defect situation can be analyzed simply and reliably by this method.

[0003] This objective is achieved by a defect analysis method in industrial process plants, for example, painting plants.

[0004] A defect analysis method in an industrial process plant, for example, a painting plant, preferably includes the following:

[0005] - Automatic recognition of defect situations, especially in industrial process plants;

[0006] - Storage of defect situation data sets for each recognized defect situation in the defect database;

[0007] - Automatic determination of the cause of the defect for each recognized defect situation and / or automatic determination of process values ​​related to the defect situation, based on the defect data set of each recognized defect situation.

[0008] In the context of this description and the attached claims, the term “process value causing a defect situation” is understood to mean, in particular, a process value that causes a defect situation and / or is associated therewith.

[0009] In the context of this description and the appended claims, the term "particularly" is used exclusively to describe possible optional and / or discretionary features.

[0010] It can be advantageous if defect situations are automatically recognized by the message system.

[0011] In one embodiment of a defect analysis method in an industrial process plant, for the purpose of automatically determining the cause of a defect in a defect situation and / or automatically determining process values ​​related to the defect situation, one or more process values ​​are provided to be automatically linked to a defect situation based on one or more of the following link criteria:

[0012] - Prior linking from the message system;

[0013] - Correlation of process values ​​with identical parts of an industrial process plant where the defect situation occurred;

[0014] - Link process values ​​to historical defect situations based on active selection by the user.

[0015] - Active selection of process values ​​by the user.

[0016] In one embodiment of a defect analysis method in an industrial process plant, for the purpose of automatically determining the cause of a defect in a defect situation and / or automatically determining process values ​​related to the defect situation, the automatic prioritization of process values ​​linked to a defect situation is characterized by being automatically performed based on one or more of the following prioritization criteria:

[0017] - Process relevance of process values;

[0018] - Location of a sensor determining a process value or a process value in an industrial process plant;

[0019] - Amount by which the process value deviates from the specified process window and / or normal conditions;

[0020] - Prioritization of history process values ​​in history defect situations;

[0021] - By adopting the prioritization of process values ​​from the defect cause and / or message system;

[0022] - Prioritization by user.

[0023] Prioritization based on the process relevance of process values ​​is preferably performed so that higher priority is given to process values ​​that are important to the process.

[0024] In the context of this description and the attached claims, the term "process value important to the process" is understood to mean a process value that is stored in a message system as important to the process and / or defined by the user as important to the process.

[0025] Prioritization based on the location of sensors or process values ​​determining process values ​​in an industrial process plant is preferably performed such that higher priority is given to process values ​​associated with the same plant component, nearby plant components, and / or comparable plant components.

[0026] In the context of this description and the attached claims, the term “comparable plant component” is understood to mean a plant component of a particularly similar or identical layout. Comparable plant components are, for example, an industrial air supply plant of the same or similar configuration, a conditioning module of an industrial air supply plant having the same or similar configuration, or a pump or motor of the same or similar configuration.

[0027] The location of the sensor determining the process value is preferably identified using a classification that includes a numbering system (so-called plant numbering system) in the industrial process plant.

[0028] Process values ​​are preferably clearly specified by a numbering system.

[0029] Preferably, process values ​​are prioritized according to their designation in the numbering system.

[0030] For the purpose of clearly specifying sensor and / or process values, the numbering system preferably includes the designation of functional units, the designation of functional groups of each functional unit, and / or the designation of functional elements of each functional group associated with each sensor and / or process value.

[0031] In addition, it may be advantageous if the clear designation of process values ​​by the numbering system includes the designation of the types of measurement variables, such as temperature, through-flow, and pressure.

[0032] For example, the air supply plant of a painting plant is a functional unit, where the conditioning module of the air supply plant is a functional group and the pump of the air supply plant is a functional element.

[0033] The normal conditions of the process value are preferably determined by an abnormality and / or defect recognition method.

[0034] Prioritization based on prioritizing historical process values ​​in a historical defect situation is preferably performed so that the process values ​​are prioritized similarly to the historical defect situation.

[0035] In one embodiment of a defect analysis method in an industrial process plant, for the purpose of automatically determining the cause of a defect for a defect situation and / or automatically determining process values ​​related to the defect situation, additional cause of defect and / or process values ​​are provided to be suggested, wherein the suggestion is automatically made based on one or more of the following suggestion criteria:

[0036] - Process relevance of process values;

[0037] - Location of a sensor determining a process value or a process value in an industrial process plant;

[0038] - Amount by which the process value deviates from the specified process window and / or normal conditions;

[0039] - Prioritization of history process values ​​in history defect situations;

[0040] - Physical dependence of process values.

[0041] It is desirable that process values ​​important to the process be proposed in a favorable manner.

[0042] Proposals based on the location of the sensor determining the process value or the process value within the industrial process plant are preferably proposed with priority to the same plant part, nearby plant part, and / or similar plant part.

[0043] Preferably, the process value is proposed according to its designation in the numbering system.

[0044] A proposal based on the prioritization of historical process values ​​in a historical defect situation is preferably made so that a high-priority process value in a historical defect situation is preferably proposed.

[0045] For the purpose of determining a proposal based on the physical dependency of the process value, the physical dependency is preferably defined by the user as a professional rule.

[0046] Preferably, the prioritization of the proposed defect causes and / or process values ​​can be modified by the user.

[0047] In one embodiment of a defect analysis method in an industrial process plant, a historical defect situation is provided to be determined from a defect database using one or more of the following similarity criteria:

[0048] - Classification of defects in historical defect situations;

[0049] - Historical defect situations of identical or comparable plant parts;

[0050] - Process value of a historical defect situation that is identical or similar to the process value of a recognized defect situation.

[0051] It is preferable that a historical defect situation having the same defect classification as the recognized defect situation be determined.

[0052] The fact that the process value of a historical defect situation is identical or similar to the process value of a recognized defect situation is preferably determined by a comparison algorithm.

[0053] In one embodiment of a defect analysis method in an industrial process plant, a historical process value identical or similar to the process value of a recognized defect situation is provided to be determined from a process database.

[0054] Preferably, a process database is searched for the purpose of determining historical process values. It may be advantageous if the fact that process values ​​are identical or similar is determined by a comparison algorithm.

[0055] Preferably, determining the history process value is performed automatically.

[0056] In one embodiment of a defect analysis method in an industrial process plant, a determined hysteretic process value is provided to be characterized as belonging to a hysteretic defect situation.

[0057] In one embodiment of a defect analysis method for an industrial process plant, for a recognized defect situation, a defect situation data set is provided to be stored in a defect database.

[0058] In one embodiment of a defect analysis method in an industrial process plant, each defect identification data set is provided to include one or more of the following defect situation data:

[0059] - Classification of defects in defect situations;

[0060] - Process value linked to a fault situation based on pre-linking from a message system;

[0061] - Information regarding the time when each defect situation occurred;

[0062] - Information regarding the period during which each defect situation occurred;

[0063] - Information regarding the location where each defect situation occurred;

[0064] - Alarm;

[0065] - Status message.

[0066] In one embodiment of a defect analysis method in an industrial process plant, a defect situation data set for each defect situation is provided to include defect identification data for clear identification of the recognized defect situation.

[0067] Preferably, defect identification data can be used for the clear designation of defect situations.

[0068] In one embodiment of a defect analysis method for an industrial process plant, documentation data and defect removal data are provided to be stored in a defect situation data set for each defect situation.

[0069] Documented data preferably includes operating instructions, manuals, circuit diagrams, procedure diagrams, and / or data sheets for plant parts affected by each defect situation.

[0070] Defect removal data preferably includes information regarding the removal of defect situations, in particular procedural guidelines for removing defect situations.

[0071] In particular, documentation data and defect removal data can also be attached to the defect situation data set by the user.

[0072] In one embodiment of a defect analysis method in an industrial process plant, process values ​​are provided to be synchronized with the defect situation that is stored and recognized during the operation of the industrial process plant.

[0073] In one embodiment of a defect analysis method in an industrial process plant, a process value is provided to have a time stamp, and the process value is configured to be clearly associated with a specific point in time by means of the time stamp.

[0074] Furthermore, the present invention relates to a defect analysis system for defect recognition in an industrial process plant, e.g., a painting plant, wherein the system takes the form for the purpose of performing the method according to the present invention in an industrial process plant, e.g., a painting plant, and is also configured for this purpose.

[0075] In addition, the present invention relates to an industrial control system comprising a defect analysis system according to the present invention.

[0076] In addition, the present invention relates to a method for predicting process deviations in an industrial process plant, for example, a painting plant.

[0077] Another objective of the present invention is to provide a method for predicting process deviations in an industrial process plant, for example, a painting plant, which can predict process deviations simply and reliably.

[0078] This objective is achieved by a method for predicting process deviations in industrial process plants, for example, painting plants.

[0079] Methods for predicting process deviations in industrial process plants, for example, painting plants, include the following:

[0080] - Automatic generation of prediction models;

[0081] - Prediction of process deviations during industrial process plant operation using a prediction model

[0082] Preferably, process deviations of process values ​​critical to production can be predicted by a prediction model.

[0083] In one embodiment of a method for predicting process deviation, the method for predicting process deviation is provided to be performed in an industrial air supply plant, a pretreatment station, a station for cathodic immersion coating and / or a drying station.

[0084] Industrial air supply plants, pretreatment stations, and / or cathodic immersion coating stations are particularly slow (sluggish) industrial process plants.

[0085] Consequently, the process values ​​critical to production in this type of industrial process method plant change very slowly only during operation.

[0086] Due to the high inertia of this type of industrial process plant, process deviations during operation can preferably be predicted at an early stage by a prediction model.

[0087] Preferably, it is therefore possible to achieve a time gain for the repair and / or maintenance of this type of industrial process plant before process deviation occurs.

[0088] An industrial air supply plant preferably includes a plurality of conditioning modules, for example, a preheating module, a cooling module, a postheating module and / or a wetting module.

[0089] Preferably, the generated prediction model can be transferred to a similar industrial process plant.

[0090] For example, one can consider that a prediction model generated for a pretreatment station can be used for a station for cathodic immersion coating.

[0091] In one embodiment of a method for predicting process deviations, process deviations of process values ​​critical to a process in an industrial process plant are provided to be predicted by a prediction model based particularly on process values ​​that change during the operation of the industrial process plant.

[0092] In the context of this description and the attached claims, the term “process value important to production” is understood to mean, in particular, a process value whose deviation from a predetermined process window results in a quality deviation, particularly a quality defect.

[0093] Process values ​​important for production in industrial air supply plants are, for example, the air temperature and relative humidity controlled by the industrial air supply plant, particularly at the exhaust section of the industrial air supply plant.

[0094] In a painting plant, air regulated by an industrial air supply plant is preferably supplied to the painting booth, and thus preferably directly affects the quality of workpieces processed in the painting booth, particularly the quality of car bodies processed in the painting booth.

[0095] For example, it is possible to predict process deviations of process values ​​important for production over a prediction range of at least about 10 minutes, for example, at least about 15 minutes, preferably at least about 20 minutes, using a prediction model.

[0096] In one embodiment of a method for predicting process deviations, process values ​​and / or state variables are stored during the operation of an industrial process plant for a predetermined period for the purpose of automatically generating a prediction model.

[0097] If the industrial process plant is an industrial air supply plant, the stored process values ​​and / or state variables preferably include the following:

[0098] - Target variables of an industrial air supply plant, specifically the air temperature and relative air humidity controlled by the industrial air supply plant, particularly at the exhaust section of the industrial air supply plant;

[0099] - Control variables, in particular, valve positions of valves in heating and / or cooling modules of industrial air supply plants, pumps, in particular the rotational speed of wetting pumps, and / or the rotational speed of ventilators;

[0100] - Internal variables, in particular, the supply and / or return flow temperatures of heating and / or cooling modules of industrial air supply plants and / or air conditions between conditioning modules;

[0101] - Measured interruption variables, particularly external temperature and / or external relative air humidity at the intake of industrial air supply plants;

[0102] - Unmeasured interruption variables; and / or

[0103] - State variables, specifically the wetting pump (on / off); manual mode for the pump (on / off); supply valve (open / closed), ventilator (on / off).

[0104] In the context of this description and the appended claims, the term "fair value" is understood to mean, in particular, a continuous time-dependent signal.

[0105] In the context of this description and the attached claims, the term "state variable" is understood to mean, in particular, discrete time-dependent events.

[0106] In one embodiment of a method for predicting process deviations, a predetermined period during which process values ​​and / or state variables are stored during the operation of an industrial process plant is provided to be predetermined based on one or more of the following criteria:

[0107] - Industrial process plants are in a ready-to-operate state for at least about 60% of a predetermined period, preferably at least about 80%, especially for production operations.

[0108] - An industrial process plant is in a production-ready state for at least about 60% of a predetermined period, preferably for at least about 80%.

[0109] - During a predetermined period, industrial process plants are operated, in particular, using all possible operating strategies;

[0110] - A predetermined number of process deviations and / or interruptions within a predetermined period.

[0111] In the case where the industrial process plant is an industrial air supply plant,

[0112] - The fan of the industrial air supply plant is operating (the fan's status variable is "on");

[0113] - The conditioning module of the industrial air supply plant operates in automatic mode;

[0114] - At least one control valve is opened; and / or

[0115] - If the wetting pump is operating (the wetting pump's state variable is "on"),

[0116] This is preferably in a state ready for operation for production operations.

[0117] In the context of this description and the attached claims, the term "production readiness of an industrial process plant" is understood to mean, in particular, that the target variable of an industrial process plant is within a predetermined process window.

[0118] If an industrial process plant is an industrial air supply plant, and the target variables of the industrial air supply plant—specifically the temperature and relative air humidity of the air controlled by the industrial air supply plant, particularly at the exhaust section of the industrial air supply plant—are within a predetermined process window, then it is in a production-ready state.

[0119] The pretreatment station or the station for cathodic immersion coating can operate specifically with only a single operating strategy.

[0120] In particular, industrial air supply plants can operate with multiple operating strategies that depend specifically on ambient conditions.

[0121] Industrial air supply plants can operate with, for example, the following operating strategies: heating / wetting, cooling / heating, cooling / wetting, cooling, heating, and wetting.

[0122] If the industrial process plant is an industrial air supply plant, it can be conceived that process values ​​and / or state variables for automatically generating a prediction model are stored, for example, for a period of at least about 6 months, particularly for a period of at least about 9 months, preferably for a period of at least about 12 months.

[0123] If the industrial process plant is a pretreatment station or a station for cathodic immersion coating, it can be conceived that process values ​​and / or state variables for automatically generating a prediction model are stored, for example, for a period of at least about 2 weeks, particularly for a period of at least about 4 weeks, preferably for a period of at least about 6 weeks.

[0124] For example, one can conceive of causing at least about 30, preferably at least about 50, process deviations and / or disruptions to occur during a predetermined period.

[0125] In particular, it can also be conceived that the predetermined period during which process values ​​and / or state variables are stored during the operation of an industrial process plant includes multiple discontinuous sub-periods.

[0126] If the period during which process values ​​and / or state variables are stored during the operation of an industrial process plant includes multiple discontinuous sub-periods, then the sub-periods preferably each have one or more of the following criteria.

[0127] - Minimum duration of the sub-period, e.g., at least about 30 minutes;

[0128] - Operation of an industrial process plant in a production-ready state and / or already in operation at the start of a sub-period;

[0129] - Operation of an industrial process plant in a state that is already in operation at the end of the sub-period.

[0130] In one embodiment of a method for predicting process deviation, a machine learning method is provided to be performed for the purpose of generating a prediction model, wherein process values ​​and / or state variables stored for a predetermined period are used to generate the prediction model.

[0131] Machine learning methods performed for the purpose of automatically generating a prediction model preferably include one or more of the following: gradient boosting, a random forest, and a support vector machine.

[0132] In one embodiment of a method for predicting process deviations, a machine learning method is provided to be performed based on features extracted from process values ​​and / or state variables stored over a predetermined period.

[0133] In one embodiment of a method for predicting process deviation, one or more of the following are provided to be used for the purpose of extracting features:

[0134] - Key statistical figures;

[0135] - Coefficients from major component analysis;

[0136] - Linear regression coefficients;

[0137] - Dominant frequency and / or amplitude from the Fourier spectrum.

[0138] Key statistical figures include, for example, minimum, maximum, median, mean, and / or standard deviation.

[0139] In one embodiment of a method for predicting process deviation, a selected number of prediction data sets having process deviation and a selected number of prediction data sets without process deviation are provided to be used to train a prediction model.

[0140] In particular, it can be considered that a selected number of prediction data sets with process deviation correspond at least approximately to a selected number of prediction data sets without process deviation.

[0141] In particular, one can consider a selected number of prediction data sets with process deviation and a selected number of prediction data sets without process deviation to be identical.

[0142] In one embodiment of a method for predicting process deviation, the selection of the number of prediction data sets having process deviation is provided to be based on one or more of the following criteria:

[0143] - Minimum time interval between two forecast data sets with process deviation;

[0144] - Automatic selection based on defined rules;

[0145] - User selection.

[0146] The minimum time interval between two prediction data sets with process deviation is, for example, at least about 2 hours.

[0147] In one embodiment of a method for predicting process deviation, a prediction data set having process deviation is provided such that the process deviation occurs within a predetermined time interval.

[0148] The predetermined time interval preferably includes the period of the prediction data set and the selected prediction range.

[0149] For example, consider a period of 30 minutes for the prediction data set and a selected prediction range of 15 minutes.

[0150] A prediction data set without process deviation is characterized as having no process deviation within a predetermined time interval.

[0151] In one embodiment of a method for predicting process deviations, process values ​​and / or state variables stored over a predetermined period are provided to be grouped into a prediction data set by preprocessing.

[0152] In one embodiment of a method for predicting process deviation, pretreatment is provided to include the following:

[0153] - Normalization of process values ​​stored over a predetermined period;

[0154] - Group process values ​​and / or state variables into a forecast data set by classifying process values ​​and / or state variables into time windows with time offsets.

[0155] Preferably, the duration of the time window is greater than the time offset.

[0156] The duration of the time window is, for example, 30 minutes.

[0157] The time offset is, for example, 5 minutes.

[0158] Preferably, each of the temporally consecutive prediction data sets here includes process values ​​and / or state variables having a time overlap of, for example, 5 minutes.

[0159] Furthermore, the present invention relates to a prediction system for predicting process deviations in an industrial process plant, wherein the prediction system takes the form for the purpose of carrying out the method according to the present invention for predicting process deviations in an industrial process plant, e.g., a painting plant, and is also configured for this purpose.

[0160] In addition, the present invention relates to an industrial control system comprising a prediction system according to the present invention.

[0161] The method according to the present invention for predicting process deviations preferably has individual or multiple features and / or advantages described together with the method according to the present invention for defect analysis.

[0162] In addition, the method according to the present invention for defect analysis preferably has individual or multiple features and / or advantages described together with the method according to the present invention for predicting process deviations.

[0163] In addition, the present invention relates to a method for recognizing abnormalities and / or defects in an industrial process plant, for example, a painting plant.

[0164] The present invention has an additional objective of providing a method for recognizing abnormalities and / or defects in an industrial process plant, e.g., a painting plant, wherein abnormalities and / or defects can be recognized simply and reliably by the method.

[0165] This objective is achieved by a method for recognizing abnormalities and / or defects in industrial process plants, for example, painting plants.

[0166] A method for recognizing anomalies and / or defects in an industrial process plant, e.g., a painting plant, preferably comprises the following:

[0167] - Automatic generation of anomaly and / or defect models of industrial process plants, including information regarding the probability of occurrence of process values;

[0168] - Automatic input of process values ​​for the industrial process plant during operation;

[0169] - Automatic recognition of abnormal and / or defect situations by determining the probability of occurrence based on the process values ​​of the input industrial process plant using an abnormal and / or defect model, and by checking the probability of occurrence against the threshold value.

[0170] Preferably, fault situations, i.e., defects and / or failures of components, sensors and / or actuators, can be identified by an anomaly and / or fault recognition method.

[0171] Preferably, the normal conditions of an industrial process plant can be determined in an automated manner by a method for recognizing abnormalities and / or defects in the industrial process plant.

[0172] In particular, the static and / or dynamic relationships of industrial process plants can be described by abnormal and / or fault models.

[0173] In the context of this description and the appended claims, the term "anomaly" is understood to mean, in particular, a deviation of process values ​​from normal conditions.

[0174] Preferably, the abnormal and / or defective model includes a structural graph.

[0175] In particular, the structure graph includes multiple cliques, wherein the relationships between the nodes of each clique are preferably described by a probability density function.

[0176] The relationship regarding sensors and / or actuators of an industrial process plant is preferably explained by each click of a structural graph.

[0177] Preferably, if the probability of occurrence of a process value within a click of the structural graph of the abnormal and / or defect model falls below a threshold value, the abnormality is recognized.

[0178] It can be advantageous if perceived anomalies with unusual process variables are graphically displayed to the user.

[0179] In one embodiment of an anomaly and / or defect recognition method,

[0180] - To ensure that the anomaly and / or defect model includes structural data containing information regarding the process structure within an industrial process plant, and / or

[0181] - Anomaly and / or defect models are provided to include parameterized data containing information regarding the relationships between process values ​​of an industrial process plant.

[0182] Structural data includes information regarding the relationships between sensors and / or actuators, in particular, of industrial process plants.

[0183] Parameterized data includes information, in particular, regarding the probability of occurrence of process values.

[0184] In particular, structural data and / or parameterization data are utilized to generate anomaly and / or defect models.

[0185] In one embodiment of an anomaly and / or defect recognition method, for the purpose of generating an anomaly and / or defect model, one or more of the following steps are provided to be performed:

[0186] - Structural identification to determine the process structure of an industrial process plant;

[0187] - Determination of causal relationships within the determined process structure of an industrial process plant;

[0188] - Structural parameterization of relationships within the determined process structure of an industrial process plant.

[0189] Anomaly and / or defect models preferably include structural information, causal relationship information, and / or structural parameterization information.

[0190] Preferably, the structure identification is configured to facilitate structure parameterization.

[0191] In particular, structure identification reduces the processing work for parameterization, and therefore specifically for structure parameterization.

[0192] In one embodiment of an anomaly and / or defect recognition method, in the context of structural identification for determining the process structure of an industrial process plant, a structural graph is provided to be determined, particularly for mapping relationships within the industrial process plant.

[0193] Preferably, the structure graph includes a plurality of nodes and a plurality of edges connecting the nodes in pairs.

[0194] Preferably, the structure graph includes a plurality of cliques.

[0195] It can be advantageous if the relationships in the determined structure graph are determined by structure identification.

[0196] In one embodiment of the method for recognizing anomalies and / or defects, the determination of a structural graph is provided to be performed using one or more of the following:

[0197] - Machine learning methods;

[0198] - Professional knowledge;

[0199] - Known circuit diagrams and / or procedure diagrams;

[0200] - Designation in the numbering system of industrial process plants.

[0201] For the purpose of structural identification, particularly for the purpose of determining structural graphs, a classification including a numbering system (so-called plant numbering system) may be advantageous if used, for example, in industrial process plants through semantic analysis.

[0202] The numbering system includes, in particular, information regarding functional units, for example, information regarding the plant type of an industrial process plant, information regarding the functional group of each functional unit, information regarding the functional element of each functional group, and / or information regarding the data type.

[0203] Preferably, the numbering system includes multiple levels.

[0204] The first level of the numbering system includes, for example, information regarding each functional unit.

[0205] The second level of the numbering system includes, for example, information regarding each function group.

[0206] The third level of the numbering system includes, for example, information regarding each functional element.

[0207] The fourth level of the numbering system includes, for example, information regarding each data type.

[0208] Preferably, the numbering system data set includes a clear designation of the functional elements of the industrial process plant.

[0209] For example, the explicit designation of a functional element includes information regarding the first, second, third, and / or fourth levels.

[0210] Preferably, in semantic analysis, information is extracted from a numbering system data set based, for example, on the clear designation of functional elements of an industrial process plant.

[0211] Preferably, in semantic analysis, one or more searches of strings in the numbering system data set are performed.

[0212] In particular, information may be provided to be extracted from the numbering system data set during this period.

[0213] During the extraction of information, a first string search is performed specifically within the numbering system data set, and the extracted data set is obtained specifically therefrom.

[0214] Information extracted from a numbering system dataset is preferably categorized for semantic analysis.

[0215] In categorization, for example, a second string search is performed on the extracted data set obtained during the extraction of information.

[0216] For example, one might consider that a specific physical variable measured by a sensor element can be identified during semantic analysis, particularly through one or more string searches.

[0217] Physical variables identifiable by semantic analysis are, for example, the following: thermodynamic variables (temperature and / or humidity); hydraulic variables (pressure, volume and / or charge level); mechanical variables (rotational speed, torque and / or rotational position); electrical variables (frequency, voltage, current intensity and / or electrical output).

[0218] Furthermore, it may be advantageous if the state variable can be identified during semantic analysis, particularly through one or more string searches.

[0219] Identifiable state variables during semantic analysis include, for example, the following information: information regarding the operating status of the wet pump (on / off); information regarding the manual mode for the pump (on / off); information regarding the operating status of the supply valve (open / closed); information regarding the operating status of the ventilator (on / off).

[0220] Preferably, determining the structure graph by a machine learning method is performed, for example, using correlation coefficients that can reproduce non-linear relationships based on mutual information.

[0221] In the context of this description and the appended claims, the term "expert knowledge" is understood to mean, for example, knowledge of the relationships between sensors in a process.

[0222] Preferably, the configuration allows edges between nodes of the structure graph to be removed by pre-configuring the structure graph with the help of expert knowledge and information from known schematics and / or procedural diagrams. In particular, the processing work for determining the structure graph can be reduced.

[0223] Process values ​​are preferably clearly specified by a numbering system ("plant numbering system").

[0224] For this reason, it may be advantageous if the structure graph is determined using the clear specification of each process value.

[0225] In particular, it is conceivable that a structural graph determined by machine learning methods is checked for validity based on professional knowledge, known circuit diagrams and / or procedure diagrams and / or designations in the numbering system of industrial process plants.

[0226] In one embodiment of an anomaly and / or defect recognition method, an industrial process plant is provided to be activated by a test signal for the purpose of structural identification, particularly to determine a structural graph.

[0227] Preferably, abnormal and / or defective situations are intentionally created during activation by a test signal.

[0228] Test signals are generated with particular consideration of technical data. In particular, limits for the test signals can be predetermined based on technical data; for example, when determining a jump function, the maximum amplitude for the control variable jump can be predetermined.

[0229] In the context of this description and the attached claims, the term "technical data" is understood to mean, in particular, one or more of the following information items:

[0230] - Sensor type (temperature sensor, throughflow sensor, valve position sensor, pressure sensor, etc.) and / or actuator type (valve, ventilator, damper, electric motor);

[0231] - Allowable value range of sensors and / or actuators;

[0232] - Signal type of sensor and / or actuator (float, integer).

[0233] In particular, industrial process plants are dynamically activated by test signals.

[0234] A test signal is a signal that can change control variables, particularly in industrial process plants. For example, the control variables of valves and / or pumps in an industrial process plant are changed by a test signal.

[0235] In one embodiment of an anomaly and / or defect recognition method, the determination of causality in a determined process structure of an industrial process plant is provided to be performed using one or more of the following:

[0236] - System input signals and system output signals generated when an industrial process plant is activated by a test signal;

[0237] - Professional knowledge;

[0238] - Known circuit diagrams and / or procedure diagrams;

[0239] - Designation in the numbering system of industrial process plants.

[0240] The causal relationships of the determined process structure are derived, for example, from system input and system output signals of an industrial process plant, and these are determined during the activation of the industrial process plant by test signals, for example, through the respective temporal progression of the system input and system output signals.

[0241] As an alternative or additionally, it can be considered that causal relationships are derived from system input signals and system output signals determined by test signals during the activation of an industrial process plant by a causal inference method.

[0242] In the context of this description and the appended claims, the term "causalities" is understood to mean the direction of causal relationships, specifically in the determined structural graph, that is, the direction indicated by the arrow.

[0243] Preferably, the process value causing the recognized anomaly can be identified through the determined process structure or the causal relationship determined in the determined structure graph.

[0244] In one embodiment of an anomaly and / or defect recognition method, for the purpose of structural parameterization of relationships within a determined process structure of an industrial process plant, one or more of the following are provided:

[0245] - Methods for determining probability density functions, particularly Gaussian mixture models;

[0246] - Known physical relationships between process values

[0247] - Physical characteristic factors of functional elements of industrial process plants, for example, characteristic factors of valves.

[0248] Preferably, structural parameterization is performed using a method for determining the probability density function, in particular using a Gaussian mixture model.

[0249] It may be advantageous if the relationship between two variables of a functional element can be explained by the physical characteristic factor diagram of the functional element of an industrial process plant.

[0250] For example, the relationship between valve position and volumetric flow rate can be explained by the valve's known valve characteristic factor diagram.

[0251] In one embodiment of an anomaly and / or defect recognition method, data from the normal operation of an industrial process plant and / or data obtained by the activation of an industrial process plant by a test signal are provided to be used for the purpose of structural parameterization, particularly using a Gaussian mixture model, utilizing a method for determining a probability density function.

[0252] For example, control, measurement, and / or moderation variables, particularly those stored in a database, are used for the purpose of structural parameterization, which uses methods to determine probability density functions.

[0253] Preferably, for the purpose of structural parameterization using a method to determine a probability density function, data from the ongoing operation of an industrial process plant is used, and this data is stored for a period of at least 2 weeks, preferably at least 4 weeks, for example, at least 8 weeks.

[0254] In one embodiment of an anomaly and / or defect recognition method, data used for structural parameterization, particularly using a Gaussian mixture model, which utilizes a method for determining a probability density function, is provided to be preprocessed before structural parameterization.

[0255] During preprocessing, preferably, data from the normal operation of the plant that is not associated with the operational readiness or production readiness state of the industrial process plant (e.g., plant off, maintenance phase, etc.) is removed through alarms and status bits describing the status of the industrial process plant.

[0256] In addition, it may be advantageous if data from the normal operation of the plant is preprocessed by filtering, for example, by a low-pass filter and / or a Butterworth filter.

[0257] Preferably, data from normal operation is interpolated at consistent time intervals.

[0258] In one embodiment of an anomaly and / or defect recognition method, a threshold value for the probability of occurrence of a process value is set during the generation of an anomaly and / or defect model, wherein an anomaly is recognized when the probability of occurrence falls below the threshold value.

[0259] The limit value is preferably set in an automated manner.

[0260] The limit value is preferably set by a non-linear optimization method, for example, by the Nelder-Mead method.

[0261] As an alternative or additionally, one can consider setting limit values ​​using quantiles.

[0262] The threshold value for the probability of occurrence of process values ​​can preferably be optimized, for example, by predetermining the false positive rate.

[0263] Preferably, after the first generation of the abnormal and / or fault model, the limit value is set, especially in the case of too many false alarms.

[0264] In one embodiment of an anomaly and / or defect recognition method, the cause of the defect in the recognized anomaly and / or recognized defect situation is provided to be identified by the anomaly and / or defect recognition method.

[0265] In particular, the cause of the defect can be identified through the structural graph of the abnormal and / or defective model.

[0266] Preferably, a structural graph for identifying abnormal and / or defective situations and / or the cause of the defect is displayed to the user.

[0267] The structural graph is configured to enable root cause analysis in particular. Specifically, abnormal process values ​​in the process structure of industrial process plants can be identified.

[0268] Preferably, the recognized abnormality can be indicated by the user as a fault situation or a false alarm.

[0269] Defect situations are stored specifically in the defect database.

[0270] In one embodiment of an anomaly and / or defect recognition method, an industrial process plant is provided to include or be formed by one or more of the following processing stations of a painting plant:

[0271] - Preprocessing station;

[0272] - Station for cathode immersion coating;

[0273] - Drying station;

[0274] - Industrial air supply plant;

[0275] - Painting robot.

[0276] Furthermore, the present invention relates to an anomaly and / or defect recognition system for recognizing anomalies and / or defects, wherein the system takes the form of performing the method according to the present invention for recognizing anomalies and / or defects in an industrial process plant, e.g., a painting plant, and is also configured to perform the method.

[0277] An anomaly and / or defect recognition system specifically forms a message system, and through this message system, defect situations within an industrial process plant can be recognized in an automated manner.

[0278] In addition, the present invention relates to an industrial control system comprising an abnormality and / or defect recognition system according to the present invention.

[0279] The method according to the present invention for recognizing anomalies and / or defects preferably has the individual or plurality of features and / or advantages described in relation to the method according to the present invention for defect analysis and / or the method according to the present invention for predicting process deviations.

[0280] The method according to the present invention for defect analysis and / or the method according to the present invention for predicting process deviations preferably has individual or multiple features and / or advantages described together with the method according to the present invention for recognizing abnormalities and / or defects.

[0281] Further features and / or advantages of the present invention are the subject of the following description and are represented in the drawings of exemplary embodiments. Brief explanation of the drawing

[0282] In the drawing, Figure 1 shows a schematic diagram of an industrial process plant and an industrial control system. Figure 2 shows a schematic diagram of an industrial process plant, in particular a painting plant. Figure 3 shows a schematic diagram of an industrial air supply plant. Figure 4 shows a schematic diagram of an industrial air supply plant from Figure 3 regarding the occurrence of a defect situation. Figure 5 shows a schematic diagram of an industrial air supply plant from Figure 3 regarding the occurrence of additional defect situations. Figure 6 shows a schematic diagram of an industrial air supply plant from Figure 3 regarding the occurrence of additional defect situations. Figure 7 shows an additional schematic diagram of an industrial air supply plant. Figure 8 shows a schematic diagram of an industrial air supply plant from Figure 7 in an operating state without process deviation. Figure 9 shows a schematic diagram of an industrial air supply plant from Figure 7 in an operating state with process deviations as a result of changing ambient conditions. Figure 10 shows a schematic diagram of an industrial air supply plant from Figure 7 in an operating state with process deviation as a result of turning on the heat recovery system. Figure 11 shows a schematic diagram of an industrial air supply plant from Figure 7 in an operating state with process deviations resulting from a valve failure. Figure 12 shows a schematic diagram of process values ​​grouped into prediction data sets. Figure 13 shows a schematic diagram of the prediction data sets from Figure 12, represented as a prediction data set with process deviation and a prediction data set without process deviation. Figure 14 shows a schematic diagram of a preprocessing station. Figure 15 shows a schematic diagram of the method steps for generating anomaly and / or defect models of a preprocessing station. Figure 16 shows a schematic diagram of a graph having a process structure derived from the pretreatment station from Figure 14. Figure 17 shows the click of a factor graph. Figure 18 shows a model of the functional relationship in the clique from Figure 17. Figure 19 shows a click corresponding to the click from Figure 17, which is extended by one node by assigning a cause of defect. Similar or functionally equivalent elements are provided with the same reference number in all drawings. Specific details for implementing the invention

[0283] Figure 1 shows an industrial control system designated as 100 overall for an industrial process plant (101).

[0284] The industrial process plant (101) is, for example, a painting plant (104), which is illustrated in Fig. 2.

[0285] In particular, a method for analyzing defects in an industrial process plant (101), particularly a painting plant (102), is described with reference to FIGS. 1 to 6.

[0286] In particular, a method for predicting process deviations in an industrial process plant (101), particularly a painting plant (102), is described with reference to FIGS. 1, FIGS. 2 and FIGS. 7 to 13.

[0287] In particular, a method for recognizing abnormalities and / or defects in an industrial process plant (101), particularly a painting plant (102), is described with reference to FIGS. 1, FIGS. 2 and FIGS. 14 to 19.

[0288] The industrial process plant (101), particularly the painting plant (102) illustrated in FIG. 2, preferably includes a plurality of processing stations (104) for processing a workpiece (106), particularly for processing a car body (108).

[0289] In the embodiment of the painting plant (102) illustrated in FIG. 2, the processing stations (104) are connected in particular in sequence and thus form a painting line (110).

[0290] For the purpose of processing the workpieces (106), particularly for the purpose of painting the vehicle body (108), the workpieces (106) preferably pass through the processing station (104) in sequence.

[0291] For example, one can consider the workpiece (106) passing through a series of processing stations (104) in the order shown.

[0292] The workpiece (106) is pretreated at the pretreatment station (112) and transported from the pretreatment station (112) to the station (114) for cathode immersion coating.

[0293] After the coating is applied to the workpiece (106), it is transported from the cathode immersion coating station (114) to the drying station (116) downstream of the cathode immersion coating station (114).

[0294] After drying the coating applied to the workpiece (106) in the cathode immersion coating station (114) at the drying station (116), the workpiece (106) is preferably transported to a base coat booth (118), where the coating is applied to the workpiece (106) once again.

[0295] After the coating is applied in the base coating booth (118), the workpiece (106) is preferably transported to the base coating drying station (120).

[0296] After drying the coating applied to the workpiece (106) in the base coating booth (118) at the base coating drying station (120), the workpiece (106) is preferably transported to a clear coat booth (122), where an additional coating is applied to the workpiece (106).

[0297] After the coating is applied in the clear coating booth (122), the workpiece (106) is preferably transported to the clear coating drying station (124).

[0298] After drying the coating applied to the workpiece (106) in the clear coating booth (122) at the clear coating drying station (124), the workpiece (106) is preferably supplied to an inspection station (126) at the end of the production process.

[0299] At the inspection station (126), quality inspection is preferably performed by a quality inspector, for example, by visual inspection.

[0300] The industrial process plant (101), particularly the painting plant (102), preferably further includes an industrial air supply plant (128) for controlling air supplied to, for example, a base coating booth (118) and / or a clear coating booth (122).

[0301] The temperature and / or relative air humidity of the air supplied to the base coating booth (118) and / or clear coating booth (122) can preferably be controlled by the industrial air supply plant (128).

[0302] The production process, preferably the painting process, can be controlled by the industrial control system (100), preferably at the processing station (104) of the industrial process plant (101), particularly the painting plant (102).

[0303] Preferably, for this purpose, the industrial control system (100) includes a process inspection system (130).

[0304] Additionally, the industrial control system (100) illustrated in FIG. 1 preferably includes a database (132).

[0305] The database (132) of the industrial control system (100) preferably includes a process database (134) and a defect database (136).

[0306] It may be more advantageous if the industrial control system (100) includes a message system (138) and an analysis system (140).

[0307] Additionally, the industrial control system (100) preferably includes a display system (142), and information can be displayed to a user by the display system.

[0308] Preferably, the display system (142) includes one or more screens on which information can be presented.

[0309] The analysis system (140) preferably includes or is formed by a defect analysis system (144).

[0310] It may be more advantageous if the message system (138) includes or is formed by a prediction system (146) for predicting process deviations in an industrial process plant (101).

[0311] Alternatively, or additionally, it may be considered that the message system (138) includes an anomaly and / or defect recognition system (148).

[0312] A defect analysis system (144) takes the form of a defect analysis method in particular in an industrial process plant (101) and is also configured to perform this method, which is described with reference to FIGS. 1 to 6.

[0313] The prediction system (146) takes the form of a method for predicting process deviations in particular in an industrial process plant (101) and is also configured to perform this method, which is described with reference to FIGS. 1, FIGS. 2 and FIGS. 7 through 13.

[0314] An anomaly and / or defect recognition system (148) is configured to perform an anomaly and / or defect recognition method in particular in an industrial process plant (101), which is described with reference to FIGS. 1, FIGS. 2 and FIGS. 14 through 19.

[0315] The industrial air supply plant (128) illustrated in FIGS. 3 to 6 preferably includes a plurality of conditioning modules (150), for example, a pre-heating module (154), a cooling module (156), a post-heating module (158) and / or a wetting module (160).

[0316] For example, the industrial air supply plant (128) of the painting plant (102) is a functional unit, where the conditioning module (150) of the air supply plant (128) is a functional group, and the circulation pump (152) of the air supply plant is a functional element (see FIGS. 3 to 6).

[0317] In addition to the circulation pumps (152) of the preheating module, cooling module (156) and postheating module (158), the industrial air supply plant (128) preferably further includes a wetting pump (153) of the wetting module (160).

[0318] It may be more advantageous if the supply plant (128) includes a ventilator (162).

[0319] The supply plant (128) preferably further includes a heat recovery system (164) for the purpose of heat recovery.

[0320] Preferably, the air stream (165) can be supplied to the air supply plant (128) from an area around the air supply plant.

[0321] An air stream (167) controlled by an air supply plant (128) can preferably be supplied to a base coating booth (118) and / or a clear coating booth (122).

[0322] Preferably, the air supply plant includes a sensor (not shown in the drawing) capable of detecting process values.

[0323] For example, the following process values ​​can be detected by a sensor, and these are preferably designated by reference numbers in FIGS. 3 to 6, respectively:

[0324] - External temperature (166);

[0325] - External humidity (168);

[0326] - Temperature of air controlled by an industrial air supply plant (170);

[0327] - Humidity of air controlled by an industrial air supply plant (172);

[0328] - Volumetric flow rate (174, 176, 178) within the conditioning module (150);

[0329] - Valve positions (180, 182, 184) of valves (181, 183, 185) within the conditioning module (150);

[0330] In addition, it may be advantageous if the rotational speed (193) of the wetting pump (153) and the rotational speed (195) of the fan (194) are detected.

[0331] Preferably, process values ​​(166 to 184) are stored in a process database (134).

[0332] Additionally, the following state variables may be provided to be detected, which are preferably likewise designated by reference numbers in FIGS. 3 to 6:

[0333] - Pump status (186, 188, 190) of the circulation pump (152) of the conditioning module (150) and pump status (192) of the wetting pump (153) (on / off);

[0334] - Valve status (194) (On / Off);

[0335] - Valve status (196, 198, 200) (open / closed) of the conditioning module (150);

[0336] - Status (202) (on / off) of the heat recovery system (164).

[0337] Preferably, the state variables (186 to 202) are also stored in the process database (134).

[0338] A method for analyzing defects in an industrial process plant (101) is now preferably described with reference to FIGS. 3 to 6.

[0339] Here, the industrial air supply plant (128) specifically forms an industrial process plant (101).

[0340] Various exemplary situations are described below, and the function of the defect analysis system (144) can be shown from these exemplary situations.

[0341] Exemplary situation 1 (see Fig. 4):

[0342] (Valve leak)

[0343] Valve leakage occurs in the preheating module (154). A volumetric flow rate (174) greater than 0 is measured. The pump state (186) is "off" and the valve state (196) of the control valve is "closed".

[0344] Fault conditions are stored in the message system as logical items (pump status (186) = off, volumetric flow rate (174) > 0, valve status (196) closed). Thus, the message system preferably stores process and status values ​​as prior links.

[0345] The defect analysis step regarding the occurrence of a message as a result of valve leakage is preferably as follows:

[0346] 1) The message is displayed to the user by the display system (142), for example, by the screen of the display system (142).

[0347] 2) The user wants to analyze the situation and opens the diagnostic window.

[0348] 3) The process value (174) and state variable (186 and 196) linked to the message are displayed to the user in the diagnostic window. Preferably, the fault analysis system (144) receives this information directly from the message system (138).

[0349] 4) Since only process values ​​(174) are associated with defect situations, process values ​​linked to defect situations are preferably not prioritized.

[0350] 5) The user saves the defect situation along with the link to the defect database (136).

[0351] 6) If a fault situation "valve leakage" occurs again in the same or similar valve, it is desirable to indicate the similar fault situation to the user.

[0352] Exemplary situation 2 (see Fig. 5):

[0353] (Excessive temperature of air (170) controlled by an industrial supply air plant (128) as a result of an excessively high external temperature (166),)

[0354] Because the external temperature (166) is outside the design window of the industrial air supply plant (128), the temperature of the air (170) controlled by the industrial air supply plant (128) is too high.

[0355] When the temperature (170) of the air controlled by the industrial air supply plant (128) deviates from a predetermined process window, a message is generated and the message system (138) transmits it to the display system (142).

[0356] The message is linked to the value of the air temperature (170) controlled by the industrial air supply plant (128). However, the message is not linked to the external temperature (166).

[0357] The defect analysis step when a message is generated as a result of a temperature change is preferably as follows:

[0358] 1) The message is displayed to the user by the display system (142), for example, by the screen of the display system (142).

[0359] 2) The user wants to analyze the situation and opens the diagnostic window.

[0360] 3) The process value (170) linked to the message is displayed to the user in the diagnostic window.

[0361] 4) Since only process values ​​(166) are associated with defect situations, process values ​​linked to defect situations are preferably not prioritized.

[0362] 5) The following process values ​​are additionally and preferably suggested to the user:

[0363] - Humidity of air controlled by an industrial air supply plant (172) (a variable important to the process, displaying abnormal behavior);

[0364] - External temperature (166) (displays abnormal behavior)

[0365] - External humidity 168 (displays abnormal behavior).

[0366] 6) The user selects a proposed process value and adds it to the defect situation.

[0367] 7) Since the relevant process value is suggested to the user, the user can directly capture the cause of the temperature deviation from the analysis system (140), especially from the defect analysis system (144).

[0368] 8) The user adds documentation to the defect situation along with suggestions for removing the defect.

[0369] 9) The user saves the defect situation, along with links and documents, to the defect database.

[0370] 10) In addition to the defect ID, defect classification (temperature rise), and defect location (exhaust section of the industrial air supply plant (128)), the defect analysis system (144) preferably also captures the references (IDs) of process variables (170, 172, 166, 168) in a prioritized order and the quantity of features (such as the mean, minimum, maximum, and scatter of the process variables during the occurrence of the defect situation), and the period from the occurrence of the defect situation to the point of storage or the end of the defect situation.

[0371] Exemplary situation 3 (see Fig. 5):

[0372] (Excessive temperature of air controlled by an industrial supply plant as a result of an excessively high external temperature (170))

[0373] Due to the external temperature, the temperature of the air (170) controlled by the industrial air supply plant is again too high. The defect pattern is similar to exemplary situation 2.

[0374] The defect analysis step when a message is generated as a result of a temperature change is preferably as follows:

[0375] 1) The message is displayed to the user by the display system (142), for example, by the screen of the display system (142).

[0376] 2) The user wants to analyze the defect situation and opens the diagnostic window.

[0377] 3) The process values ​​(170, 172, 166, 168) linked in the message are displayed to the user in the diagnostic window in the order shown.

[0378] 4) A list of processes and their prioritization are generated from similar defect situations. Similarity to the defect situation from exemplary situation 2 is determined by the defect analysis system (144) by metric matching of process values.

[0379] 5) A similar defect situation is displayed to the user.

[0380] 6) To find a solution, the user utilizes the trends of process values ​​in the displayed prioritized order and documents of similar defect situations displayed.

[0381] 7) The user saves the defect situation.

[0382] Exemplary situation 4 (see Fig. 6) :

[0383] (Disruption of supply system)

[0384] Too little combustion gas is supplied to the burner in the preheating module (154). The volumetric flow rate (174) is reduced.

[0385] To accommodate more combustion gas, the valve (181) of the preheating module (154) is opened further, and the valve position (180) is changed.

[0386] To compensate for the interruption of the preheating module (154), the valve (185) of the postheating module (158) is also opened. The valve position (184) is changed.

[0387] Due to the low external temperature (166), the interruption cannot be compensated for, and the temperature of the air (170) controlled through the industrial supply plant drops sharply.

[0388] The steps when a message occurs as a result of a supply system interruption are preferably as follows:

[0389] 1) The message is displayed to the user by the display system (142), for example, using the screen of the display system (142).

[0390] 2) The user opens the diagnostic window hoping to analyze the defect situation.

[0391] 3) As a result of prioritization in the message system (138), the process value (170) linked to the message is displayed to the user in the diagnostic window.

[0392] 4) Prioritization does not occur.

[0393] 5) The following process values ​​are proposed:

[0394] - Humidity of air controlled by an industrial air supply plant (172) (a variable important to the process, displaying abnormal behavior);

[0395] - Volumetric flow rate (174), valve position (180), valve position (186) (depends on deviation from normal conditions).

[0396] 6) The defect situations from exemplary situations 2 and 3 are not classified as similar (different signal behavior due to the high metric distance between process values).

[0397] 7) The proposed process value can be added to the defect situation and stored in the defect database.

[0398] Now, a method for predicting process deviations in an industrial process plant (101) is preferably described with reference to FIGS. 7 to 13.

[0399] If the industrial process plant (101) is an industrial air supply plant (128), the stored process values ​​and / or state variables preferably include the following (see FIG. 7):

[0400] - Target variables (204) of an industrial air supply plant (128), particularly air temperature (170) and relative air humidity (172) controlled by the industrial air supply plant (128), particularly at the exhaust section of the industrial air supply plant (128);

[0401] - Control variables (206), particularly valve positions (180, 182, 184) of the valves of the heating and / or cooling modules (154, 156, 158) of the industrial air supply plant (128), rotational speed (193) of the pump (152), particularly the wetting pump (153), and / or rotational speed (195) of the fan (162);

[0402] - Internal variables (208), in particular, the supply and / or return flow temperatures (210) of the heating and / or cooling modules (154, 156, 158) of the industrial air supply plant (128) and / or air conditions between the conditioning modules (150);

[0403] - Measured interruption variables (210), particularly external temperature (166) and / or external relative air humidity (168) at the intake of an industrial air supply plant (128);

[0404] - Unmeasured interruption variable (212); and / or

[0405] - State variable (214), in particular, wetting pump (153) (on / off); manual mode (on / off) for pump (152, 153); supply valve (181, 183, 185), ventilator (162) (on / off).

[0406] Various exemplary operating states in which the function of the prediction system (146) can be shown are described below.

[0407] Exemplary operating state 1:

[0408] (No deviation)

[0409] FIG. 8 shows a first exemplary operating state of an industrial air supply plant (128) without process deviation. Thus, the first exemplary operating state preferably represents a positive case.

[0410] The external temperature (166) and external humidity (168) are not constant. The preheating module (154) and the wetting module (160) are active.

[0411] The control system of the industrial air supply plant (128) maintains the temperature (170) and relative air humidity (172) of the air controlled by the industrial air supply plant (128) at a constant value.

[0412] According to the state variable (214) (ventilator (162) = on and valve and pump mode = automatic), the industrial air supply plant (128) is ready for operation.

[0413] Additionally, due to the constant temperature (170) and relative air humidity (172) of the air controlled by the industrial air supply plant (128), the industrial air supply plant (128) is preferably production-ready.

[0414] Exemplary operating state 2:

[0415] (As the external relative humidity (168) drops, the external temperature (166) above the technical parameter; the effect of the measured interruption variable (210) on the target variable (204)

[0416] FIG. 9 shows a second exemplary operating state of an industrial air supply plant (128) in which the external relative humidity (168) drops and the external temperature (166) increases as a result of a sudden change in weather.

[0417] The process value displays the following behavior:

[0418] a) Increased external temperature (166) after a sudden change in weather is outside the technical parameters.

[0419] b) The power of the previous active preheater module (154) is reduced by the controller.

[0420] c) Since the reduction in heating power is insufficient, the controller opens the valve (183) of the cooling module (156) and thus increases the cooling power.

[0421] d) To compensate for falling external humidity (168), the rotational speed (193) of the wetting pump (153) is increased by the controller.

[0422] e) As a result of the insufficient design of the cooling module (156), there is a deviation in the temperature (170) of the air controlled by the industrial air supply plant (128) because the cooling power is insufficient.

[0423] Departure from a predetermined process window for temperature (170) is delayed as a result of the inertia of the industrial air supply plant (128) and compensation by the controller.

[0424] Exemplary operating condition 3 :

[0425] (Switching to winter operation with heat recovery; effect of unmeasured interruption variable (212) on target variable (204))

[0426] FIG. 10 shows a third exemplary operating state of an industrial air supply plant (128) when a heat recovery system (164) that uses waste heat to heat the airflow (165) in cold climate conditions is turned on.

[0427] The heat recovery system (164) is turned on by a manual valve, and for this reason, the effect of heat recovery by the heat recovery system (164) is not measurable (unmeasurable interruption variable (212)).

[0428] The process value displays the following behavior:

[0429] a) The heating power of the heat recovery system (164) is increased; the value is not measurable.

[0430] b) The valve (181) of the preheating module (154) is closed due to an increase in heating power.

[0431] c) In the cooling module (156), the valve (183) is opened to maintain the temperature by additional cooling power.

[0432] d) The rotational speed (193) of the humidification pump (153) is adjusted by a controller so that the air humidity (172) of the air controlled by the industrial air supply plant (128) is maintained.

[0433] e) There is a deviation in the temperature (170) of the air controlled by the industrial air supply plant (128) because the cooling module (156) cannot quickly and sufficiently compensate for the heat supply. The deviation occurs with a delay due to the inertia of the industrial air supply plant (128) and compensation by the controller.

[0434] Exemplary operating condition 4 :

[0435] (Failure of the valve (181) of the preheating module (154))

[0436] FIG. 11 shows a fourth exemplary operating state of an industrial air supply plant (128) when there is a failure of the valve (181) of the preheating module (154).

[0437] The process value displays the following behavior:

[0438] a) The valve (181) of the preheating module (154) is closed due to a device defect, and as a result, the heating power is reduced.

[0439] b) To compensate for insufficient heating power, the valve (185) of the rear heating module (158) is opened.

[0440] c) Because the heating power of the rear heating module (158) is insufficient, there is a deviation in the temperature (170) of the air controlled by the industrial air supply plant (128).

[0441] A method for predicting process deviations in an industrial process plant (101), particularly an industrial air supply plant, is described below with reference to operating states 1 to 4 described above.

[0442] Preferably, operating conditions (2 to 4) having process deviations during the operation of an industrial air supply plant (128) can be predicted by a method of predicting process deviations, for example, with a prediction horizon (216) of about 15 minutes.

[0443] As a data basis for training a prediction model, the period during which an industrial air supply plant (128) operates in normal operation (>80%) from an operating state ready for use (see exemplary operating state 1) is considered.

[0444] The recorded data preferably includes exemplary operating states 2 to 4 multiple times in each case. This could occur during operation in progress, or alternatively, could be intentionally caused by, for example, closing valves (181, 183, 185).

[0445] The data is preferably preprocessed and normalized, as can be shown in FIG. 12, for example. Therein, the valve position (180) of the valve (181) of the preheating module (154) and the air temperature (170) and relative air humidity (172) controlled by the industrial air supply plant (128) are recorded as examples.

[0446] The normalized data is divided into, for example, 30-minute time windows (218), with, for example, a 5-minute time offset in each case.

[0447] Data normalized by a time window (218) forms, in particular, a prediction data set, specifically a prediction data set (220) without process deviation and a prediction data set (222) with process deviation (see FIG. 7).

[0448] For a prediction data set (222) having process deviations, a state variable (214) is used to check whether the industrial process plant (101) is ready for operation (e.g., a blower (162) in the ON state, a conditioning module (150) in the automatic mode state).

[0449] - Otherwise: The corresponding prediction data set (222) with process deviation is rejected and is not used to train the prediction model.

[0450] - Then: The corresponding prediction data set (222) with process deviation is a contender for training the prediction model.

[0451] For example, due to the minimum time interval of 1 hour, only one prediction data set (222) with process deviation in FIG. 7 is selected.

[0452] Preferably, the selection of a prediction data set (220) without process deviation is performed similarly to the selection of a prediction data set (222) with process deviation.

[0453] For example, with a minimum time interval of 1 hour, only one prediction data set (220) without process deviation is selected to train the prediction model.

[0454] After that, features are preferably extracted from a selected prediction data set (220) without process deviation and a selected prediction data set (222) with process deviation.

[0455] For the purpose of extracting features, for example, key statistical figures, such as minimum, maximum, median, mean, and / or standard deviation, are used. It may be more advantageous if linear regression coefficients are used to extract features.

[0456] Preferably, the prediction model is trained based on features extracted from a selected prediction data set (220) without process deviation and based on a selected prediction data set (222) with process deviation, particularly by a machine learning method, for example, gradient boosting.

[0457] Using a trained prediction model, process deviations of process values ​​important for production in an industrial air supply plant (128) are preferably predicted based on process values ​​that change during the operation of the industrial air supply plant (128).

[0458] In particular, the predictive model is described with reference to exemplary operating states 2 through 4:

[0459] Exemplary operating state 2:

[0460] The prediction model predicts the process deviation after the occurrence of a temperature rise. The basis for this is the course of the measured interruption variables (210), in particular the external temperature (166) and external humidity (168), the response of the conditioning module (150), and the temperature (170) of the air controlled by the industrial supply plant (128) in the exhaust section.

[0461] Exemplary operating condition 3 :

[0462] The prediction model predicts an increase in the temperature (170) of the air regulated by the industrial supply air plant (128) after the heat recovery system is turned on. The basis is the response of the conditioning module (150) and the trend of the temperature (170) of the air regulated by the industrial supply air plant (128) at the exhaust section.

[0463] Exemplary operating state 4:

[0464] The prediction model predicts an increase in the temperature (170) of the air controlled by the industrial air supply plant (128) based on weather conditions and the valve position (180) of the valve (181) of the preheating module (156).

[0465] Now, a method for recognizing abnormalities and / or defects in an industrial process plant (101) is preferably described with reference to FIGS. 14 to 19.

[0466] Fault situations, particularly deficiencies and / or defects in components, sensors and / or actuators, can preferably be identified by an abnormality and / or defect recognition method.

[0467] Here, the pretreatment station (112) forms, for example, an industrial-process plant (101).

[0468] Preferably, the pretreatment station (112) includes a pretreatment tank (224) capable of pretreating a workpiece (106), preferably a vehicle body (108).

[0469] Preferably, the pretreatment station (112) further includes a first pump (226), a second pump (228), a heat exchanger (230), and a valve (232).

[0470] Fair value ( , S 86 , T 95 , T 85 , T 15 and T 05 ) is given his designation based on the clear designation in the numbering system of the industrial process plant (101).

[0471] Process value (T 95 , T 85 , T 15 and T 05 ) indicates the temperature, particularly within the industrial process plant (101), especially within the pretreatment station (112).

[0472] Process value (S 86 ) is the valve position of the valve (232).

[0473] Fair value ( ) is the volumetric flow rate.

[0474] Preferably, for the purpose of performing an anomaly and / or defect recognition method, an anomaly and / or defect model (233) of an industrial process plant (101), particularly a pretreatment station (112), is generated, and this model includes information regarding the probability of occurrence of the process value mentioned above (see FIG. 15).

[0475] The abnormal and / or defect model (233) is preferably generated as follows:

[0476] First, a test signal is generated by considering technical data (234), particularly in the context of test signal generation (236).

[0477] In particular, the limit for the test signal is determined based on the maximum amplitude of the control variable jump when determining technical data (234), for example, the jump function.

[0478] Technical data (234) includes, for example, one or more of the following information items:

[0479] - Sensor type (temperature sensor, through-flow sensor, valve position sensor, pressure sensor, etc.) and / or actuator type (valve, ventilator, damper, electric motor);

[0480] - Allowable value range of sensors and / or actuators;

[0481] - Signal type of sensor and / or actuator (float, integer).

[0482] An industrial process plant (101), particularly a pretreatment station (112), is preferably dynamically operated by a test signal. This is indicated by reference number 238 in FIG. 15. Here, it can be conceived that abnormal and / or fault conditions are intentionally created when operated by the test signal.

[0483] During the activation of an industrial process plant (101), particularly a pretreatment station (112), a test signal, preferably a system input signal (240) and a system output signal (242), is generated.

[0484] The system input signal (240) and the system output signal (242) are preferably stored in a test signal database (244).

[0485] After that, structure identification (246) of the industrial process plant (101), particularly the pretreatment station (112), is preferably performed. Here, preferably, a structure graph (247) of the industrial process plant (101), particularly the pretreatment station (112), is determined (see FIG. 16).

[0486] Structure identification (246), particularly the determination of the structure graph, is preferably performed using a machine learning method, preferably using a correlation coefficient that can reproduce a non-linear relationship by mutual information, for example.

[0487] For the purpose of structural identification (246), it may be more advantageous if specialized knowledge (248), that is, specific knowledge of the relationship in the process, is used.

[0488] Here, for example, edges between nodes of the structure graph to be determined can be removed by pre-configuring the structure graph with the help of professional knowledge and information from known circuit diagrams and / or procedure diagrams (250). In particular, here, processing work for determining the structure graph can be reduced.

[0489] It may be more advantageous if the structural graph is determined by using the clear designation of process values ​​through the numbering system of the industrial process plant (101), particularly the pre-processing station (112), that is, by using the semantics (252) of the designation of process values.

[0490] In particular, it is conceivable that a structure graph determined by a machine learning method is checked for validity by professional knowledge (248), known circuit diagrams and / or procedure diagrams (250) and / or designation in the numbering system of an industrial process plant (101) (semantics 252).

[0491] Preferably, the causalities (254) of the determined process structure, in particular the direction indicated by the arrow in the determined structure graph, are then determined.

[0492] The causal relationship (254) of the determined process structure is derived, for example, from the system input signal (240) and system output signal (242) of the industrial process plant (101), and is determined during the activation of the industrial process plant (101) by the test signal, for example, through the temporal progression of the system input signal (240) and the system output signal (242).

[0493] Alternatively, or additionally, it can be conceived that a causal relationship (254) is derived from a system input signal (240) and a system output signal (242) determined by a test signal during the activation of an industrial process plant (101) by a causal inference method.

[0494] To determine the causal relationship (254), more preferably, there is professional knowledge (248), a procedure diagram (250), and / or a designation (semantics) (252) in the numbering system of the industrial process plant (101).

[0495] Preferably, the process value causing the recognized abnormality can be found in the determined process structure or by the determined causal relationship (254) in the determined structure graph.

[0496] After determining the structure identification (246) and / or causal relationship (254), preferably, structure parameterization (256) is performed.

[0497] Preferably, the structure identification (246) is configured to facilitate the structure parameterization (256). Preferably, the structure identification (246) is configured to reduce the processing work for the structure parameterization (256).

[0498] Preferably, structural parameterization (256) is performed using a method for determining a probability density function, in particular using a Gaussian mixture model.

[0499] Structural parameterization (256) is performed on the common probability density function (f1) of the clique (258) shown in FIG. 17, for example, using a Gaussian mixture model (see FIG. 18).

[0500] Preferably, professional knowledge (248) is likewise used for structural parameterization (256).

[0501] For the purpose of structural parameterization (256), known physical relationships between process values ​​and / or physical characteristic factor diagrams of functional elements of an industrial process plant (101), particularly a pretreatment station, are used. For example, a characteristic factor diagram of a valve (232) is used.

[0502] It may be more advantageous if professional knowledge (248) regarding defect situations is used for structural parameterization (256).

[0503] For example, valve position (S 86 ) and volumetric flow rate( The relationship between ) can be explained by the known valve characteristic factor diagram of the valve (232).

[0504] Data stored in the operation database (260) and / or data from the test signal database (244) from the normal operation of the industrial process plant (101), particularly the preprocessing station (112), are preferably used for the purpose of structural parameterization (256) using a method for determining a probability density function, particularly using a Gaussian mixture model.

[0505] For example, control, measurement, and / or control variables stored in the database (244, 260) are used for the purpose of structural parameterization (256) using a method to determine the probability density function.

[0506] Preferably, for the purpose of structural parameterization (256) using a method for determining a probability density function, data from the ongoing operation of an industrial process plant (101), particularly a preprocessing station, is used, and this data is stored for a period of at least 2 weeks, preferably at least 4 weeks, for example, at least 8 weeks.

[0507] The data is preferably preprocessed before structural parameterization (256).

[0508] During preprocessing, data from the industrial process plant (101) (e.g., plant off, maintenance phase, etc.) that is not preferably associated with the operation-ready or production-ready operating state of the industrial process plant (101) is removed through alarms and status bits describing the status of the industrial process plant (101), particularly the preprocessing station (112).

[0509] Additionally, it may be advantageous if data from an industrial process plant (101) is preprocessed by filtering, for example, by a low-pass filter and / or a Butterworth filter.

[0510] Preferably, the data is further interpolated at consistent time intervals.

[0511] During the generation of the abnormal and / or defect model (233), the limit value for the probability of occurrence of the process value is preferably set in the context of limit value optimization (264).

[0512] The threshold value for the probability of occurrence is preferably set so that an anomaly is recognized when the probability of occurrence falls below the threshold value.

[0513] The limit value is preferably set by a non-linear optimization method, for example, by the Nelder-Mead method.

[0514] As an alternative or additionally, one can consider setting limit values ​​using quantiles.

[0515] The threshold value for the probability of occurrence of a process value can preferably be optimized, for example, by predetermining the false-positive rate.

[0516] In addition, especially in the case of too many false alarms, one can consider that the limit value is adjusted after the primary generation of the abnormal and / or fault model (233).

[0517] Preferably, abnormality and / or defect recognition is performed using an abnormality and / or defect model (233) as follows.

[0518] For example, the valve (232) experiences a valve failure, and therefore the sensor value deviates from the mapped normal conditions within the individual click.

[0519] During the operation of an industrial process plant (101), particularly a pretreatment station (112), the probability of occurrence of sensor values ​​within the clock is evaluated, and if they fall below a calculated limit value, anomalies are detected in different clocks.

[0520] A valve failure of the valve (232) initially causes an abnormality in the click (258) of the valve position (S86), where a message is output by the abnormality and / or fault recognition system (148).

[0521] Over time, additional abnormalities are generated as a result of defect propagation, which later affects process-related variables, such as the tank temperature (T35) of the tank (224).

[0522] Preferably, the message from the anomaly and / or defect recognition system (148) includes one or more of the following information items:

[0523] - Time when an anomaly was detected;

[0524] - Click(s) where an anomaly affecting the sensor has occurred.

[0525] As a result of the above early recognition and message to the user, it is preferably possible to prevent deviations in process-related variables, namely the tank temperature (T35) of the tank (224), through rapid intervention.

[0526] The user can subsequently identify the cause of the defect (i.e., valve failure) for the occurrence of the above.

[0527] As a result of assigning the cause of the defect, the click (258) is extended by one node (266), and the probability density function of the anomalous data is integrated into the functional relationship (see FIG. 19).

[0528] After integrating the defect causes, the anomaly and / or defect recognition method is performed as described above. If an anomaly occurs, the probability of the defined defect cause is additionally output.

[0529] As a result of a message from the abnormal and / or defect recognition system (148), the user now receives one or more of the following information items:

[0530] - Time when an anomaly was detected;

[0531] - Click(s) where an anomaly has occurred affecting the sensor;

[0532] - Probability of the defined cause of the defect.

[0533] Specific embodiments are as follows:

[0534] Example 1:

[0535] A method for analyzing defects in an industrial process plant (101), particularly a painting plant (102), includes the following:

[0536] - In particular, automatic recognition of defect conditions in industrial process plants (101);

[0537] - Storage of a defect situation data set for each recognized defect situation in the defect database (136);

[0538] - Automatic determination of the cause of the defect for each recognized defect situation and / or automatic determination of process values ​​related to the defect situation, based on the defect data set of each recognized defect situation.

[0539] Example 2:

[0540] The method according to Example 1 is characterized in that, for the purpose of automatically determining the cause of a defect in a defect situation and / or automatically determining process values ​​related to the defect situation, one or more process values ​​are automatically linked to a defect situation based on one or more of the following link criteria.

[0541] - Pre-linking from the message system;

[0542] - Association of process values ​​with the same part of the industrial process plant (101) where the defect situation occurred;

[0543] - Link process values ​​to historical defect situations based on active selection by the user.

[0544] - Active selection of process values ​​by the user.

[0545] Example 3:

[0546] The method according to Example 2 is characterized in that, for the purpose of automatically determining the cause of a defect in a defect situation and / or automatically determining the process value associated with the defect situation, the automatic prioritization of the process value linked to the defect situation is automatically performed based on one or more of the following prioritization criteria:

[0547] - Process relevance of process values;

[0548] - Location of a sensor that determines the process value or the process value in an industrial process plant (101);

[0549] - Amount by which the process value deviates from the specified process window and / or normal conditions;

[0550] - Prioritization of history process values ​​in history defect situations;

[0551] - By adopting the prioritization of process values ​​from the defect cause and / or message system (138);

[0552] - Prioritization by user.

[0553] Example 4:

[0554] A method according to any one of Examples 1 to 3 is characterized in that additional defect causes and / or process values ​​are proposed for the purpose of automatically determining a defect cause for a defect situation and / or automatically determining a process value related to a defect situation, wherein the proposal is automatically made based on one or more of the following proposal criteria:

[0555] - Process relevance of process values;

[0556] - Location of a sensor that determines the process value or the process value in an industrial process plant (101);

[0557] - Amount by which the process value deviates from the specified process window and / or normal conditions;

[0558] - Prioritization of history process values ​​in history defect situations;

[0559] - Physical dependence of process values.

[0560] Example 5:

[0561] A method according to any one of Examples 1 to 4 is characterized in that a historical defect situation is determined from a defect database (136) using one or more of the following similarity criteria:

[0562] - Classification of defects in historical defect situations;

[0563] - Historical defect situations of identical or comparable plant parts;

[0564] - Process value of a historical defect situation that is identical or similar to the process value of a recognized defect situation.

[0565] Example 6:

[0566] A method according to any one of Examples 1 to 5 is characterized in that a historical process value identical or similar to the process value of a recognized defect situation is determined from a process database (134).

[0567] Example 7:

[0568] The method according to Example 6 is characterized in that the determined hysteresis process value is characterized as belonging to a hysteresis defect situation.

[0569] Example 8:

[0570] A method according to one of Examples 1 to 7 is characterized in that, for a recognized defect situation, a defect situation data set is stored in a defect database (136).

[0571] Example 9:

[0572] The method according to Example 8 is characterized in that each defect identification data set includes one or more of the following defect situation data:

[0573] - Classification of defects in defect situations;

[0574] - Process value linked to a fault situation based on pre-linking from a message system;

[0575] - Information regarding the time when each defect situation occurred;

[0576] - Information regarding the period during which each defect situation occurred;

[0577] - Information regarding the location where each defect situation occurred;

[0578] - Alarm;

[0579] - Status message

[0580] Example 10:

[0581] The method according to Example 8 or 9 is characterized in that the defect situation data set of each defect situation includes defect identification data for clear identification of the recognized defect situation.

[0582] Example 11:

[0583] A method according to one of Examples 8 to 10 is characterized in that documentation data and defect removal data are stored in a defect situation data set of each defect situation.

[0584] Example 12:

[0585] A method according to one of Examples 8 to 11 is characterized in that the process value is stored during the operation of the industrial process plant (101) and synchronized with the recognized defect situation.

[0586] Example 13:

[0587] A method according to any one of Examples 8 to 12 is characterized in that the process value has a time stamp, and the process value is configured to be clearly associated with a specific point in time by the time stamp.

[0588] Example 14:

[0589] As a defect analysis system (144) for defect analysis in an industrial process plant (101), for example, a painting plant (102), the system takes the form for the purpose of performing a defect analysis method in an industrial process plant (101), for example, a painting plant (102), according to one of embodiments 1 to 13, and is also configured for this purpose.

[0590] Example 15:

[0591] The industrial control system (100) includes a defect analysis system (144) according to Example 14.

[0592] Example 16:

[0593] A method for predicting process deviation in an industrial process plant (101), for example, a painting plant (102), includes the following:

[0594] - Automatic generation of prediction models;

[0595] - Prediction of process deviation during operation of industrial process plant (101) using a prediction model.

[0596] Example 17:

[0597] The method according to Example 16 is a method for predicting process deviations and is characterized by being performed in an industrial air supply plant (128), a pretreatment station (112), a station for cathode immersion coating (114) and / or a drying station (116, 120, 124).

[0598] Example 18:

[0599] The method according to Example 16 or 17 is characterized in that a process deviation of a process value important for production in an industrial process plant (101) is predicted by a prediction model based on a changing process value during the operation of the industrial process plant (101).

[0600] Example 19:

[0601] A method according to one of Examples 16 to 18 is characterized in that process values ​​and / or state variables are stored during the operation of an industrial process plant (101) for a predetermined period for the purpose of automatically generating a prediction model.

[0602] Example 20:

[0603] The method according to Example 19 is characterized in that a predetermined period during which process values ​​and / or state variables are stored during the operation of an industrial process plant (101) is predetermined based on one or more of the following criteria:

[0604] - The industrial process plant (101) is in a ready-to-operate state for at least about 60% of a predetermined period, preferably at least about 80%, particularly for production operation;

[0605] - The industrial process plant (101) is in a production ready state for at least about 60% of a predetermined period, preferably for at least about 80%;

[0606] - During a predetermined period, the industrial process plant (101) is operated using all possible operating strategies, in particular;

[0607] - A predetermined number of process deviations and / or interruptions within a predetermined period.

[0608] Example 21:

[0609] The method according to Example 19 or 20 is characterized in that a machine learning method is performed for the purpose of generating a prediction model, wherein process values ​​and / or state variables stored for a predetermined period are used to generate the prediction model.

[0610] Example 22:

[0611] The method according to Example 21 is characterized in that the machine learning method is performed based on features extracted from process values ​​and / or state variables stored over a predetermined period.

[0612] Example 23:

[0613] The method according to Example 22 is characterized in that one or more of the following are used for the purpose of extracting features:

[0614] - Key statistical figures;

[0615] - Coefficients from major component analysis;

[0616] - Linear regression coefficients;

[0617] - Dominant frequency and / or amplitude from the Fourier spectrum.

[0618] Example 24:

[0619] A method according to one of Examples 16 to 23 is characterized in that a selected number of prediction data sets (222) having process deviation and a selected number of prediction data sets (220) without process deviation are used to train a prediction model.

[0620] Example 25:

[0621] The method according to Example 24 is characterized in that the selection of the number of prediction data sets having process deviations is based on one or more of the following criteria:

[0622] - Minimum time interval between two forecast data sets with process deviation;

[0623] - Automatic selection based on defined rules;

[0624] - User selection.

[0625] Example 26:

[0626] The method according to Example 24 or 25 is characterized in that a prediction data set having process deviation is characterized such that the process deviation occurs within a predetermined time interval.

[0627] Example 27:

[0628] The method according to Example 26 is characterized in that process values ​​and / or state variables stored for a predetermined period are grouped into a prediction data set by preprocessing.

[0629] Example 28:

[0630] The method according to Example 27 is characterized in that the pretreatment includes the following:

[0631] - Normalization of process values ​​stored over a predetermined period;

[0632] - Group process values ​​and / or state variables into a forecast data set by classifying process values ​​and / or state variables into time windows with time offsets.

[0633] Example 29:

[0634] As a prediction system (146) for predicting process deviations in an industrial process plant, the prediction system takes the form for the purpose of performing a method to predict process deviations in an industrial process plant (101), e.g., a painting plant (102), according to one of Examples 16 to 29, and is also configured for this purpose.

[0635] Example 30:

[0636] The industrial control system (100) includes a prediction system (146) according to Example 29.

[0637] Example 31:

[0638] A method for recognizing abnormalities and / or defects in an industrial process plant (101), e.g., a painting plant (102), and the method comprises the following:

[0639] - Automatic generation of an abnormality and / or defect model (233) of an industrial process plant (101) including information regarding the probability of occurrence of process values;

[0640] - Automatic input of process values ​​of the industrial process plant (101) during operation;

[0641] - Automatic recognition of abnormal and / or defect situations by determining the probability of occurrence by the abnormal and / or defect model (233) based on the process value of the input industrial process plant (101) and by checking the probability of occurrence for the limit value.

[0642] Example 32:

[0643] The method according to Example 31 is,

[0644] - The abnormal and / or defect model (233) includes structural data containing information regarding the process structure of the industrial process plant (101) and / or

[0645] - The abnormal and / or defect model (233) is characterized by including parameterized data that includes information regarding the relationship between process values ​​of the industrial process plant (101).

[0646] Example 33:

[0647] The method according to Example 31 or 32 is characterized in that one or more of the following steps are performed for the purpose of generating an abnormality and / or defect model (233):

[0648] - Structure identification (246) for determining the process structure of an industrial process plant (101);

[0649] - Determination of causal relationships (254) within the determined process structure of an industrial process plant (101);

[0650] - Structural parameterization (256) of relationships within the determined process structure of an industrial process plant (101).

[0651] Example 34:

[0652] The method according to Example 33 is characterized in that, in the context of structure identification (246) for determining the process structure of an industrial process plant (101), a structure graph mapping relationships in particular in the industrial process plant (101) is determined.

[0653] Example 35:

[0654] The method according to Example 34 is characterized in that the determination of the structure graph is performed using one or more of the following items.

[0655] - Machine learning methods;

[0656] - Expertise (248);

[0657] - Notified circuit diagram and / or procedure diagram (250);

[0658] - Designation in the numbering system of an industrial process plant (101).

[0659] Example 36:

[0660] The method according to Examples 33 to 35 is characterized in that an industrial process plant (101) is activated by a test signal for the purpose of structural identification, particularly to determine a structural graph.

[0661] Example 37:

[0662] A method according to one of Examples 33 to 36 is characterized in that the determination of the causal relationship (254) in the determined process structure of the industrial process plant (101) is performed using one or more of the following items.

[0663] - System input signal (240) and system output signal (242) generated when the industrial process plant (101) is activated by a test signal;

[0664] - Expertise (248);

[0665] - Known circuit diagrams and / or procedure diagrams (252);

[0666] - Designation in the numbering system of an industrial process plant (101).

[0667] Example 38:

[0668] A method according to any one of Examples 33 to 37 is characterized in that one or more of the following are used for the purpose of structural parameterization (246) of a relationship in a determined process structure of an industrial process plant (101):

[0669] - Methods for determining probability density functions, particularly Gaussian mixture models;

[0670] - The known physical relationship between process values;

[0671] - Physical characteristic factors of the functional elements of the industrial process plant (101), for example, characteristic factors of the valve (232).

[0672] Example 39:

[0673] The method according to Example 38 is characterized in that it is used for the purpose of structural parameterization (246) using a Gaussian mixture model, in particular, using data obtained from normal operation of the industrial process plant (101) and / or activation of the industrial process plant (101) by a test signal to determine a probability density function.

[0674] Example 40:

[0675] The method according to Example 39 is characterized in that the data used for structural parameterization (246), which uses a method for determining a probability density function, particularly a Gaussian mixture model, is preprocessed before structural parameterization (246).

[0676] Example 41:

[0677] A method according to one of Examples 31 to 40 is characterized in that, during the generation of an anomaly and / or defect model (233), a limit value for the probability of occurrence of a process value is set, and if this falls below the limit value, an anomaly is recognized.

[0678] Example 42:

[0679] A method according to any one of Examples 31 to 41 is characterized in that the cause of a defect in a recognized abnormality and / or recognized defect situation is identified by an abnormality and / or defect recognition method.

[0680] Example 43

[0681] A method according to one of Examples 31 to 42 is characterized in that the industrial process plant (101) includes or is formed by one or more of the next processing stations (104) of the painting plant.

[0682] - Preprocessing station (112)

[0683] - Station for cathode immersion coating (114);

[0684] - Drying Station(116, 120, 124);

[0685] - Industrial air supply plant (128);

[0686] - Painting robot.

[0687] Example 44:

[0688] An anomaly and / or defect recognition system (148) for recognizing anomalies and / or defects takes the form of an anomaly and / or defect recognition method in an industrial process plant (101), e.g., a painting plant (102), according to one of Examples 31 to 43, and is also configured to perform this method.

[0689] Example 45:

[0690] The industrial control system (100) includes an abnormality and / or defect recognition system (148) according to Example 44.

[0691] Example 46:

[0692] The industrial control system includes a defect analysis system according to Example 14, a prediction system for predicting process deviations in an industrial process plant according to Example 29, and / or an anomaly and / or defect recognition system according to Example 44.

Claims

Claim 1 A method for analyzing defects in an industrial process plant (101), comprising: automatically recognizing a defect situation in the industrial process plant (101); storing a defect situation data set for each recognized defect situation in a defect database (136); and steps of automatically determining the cause of the defect for the defect situation and / or automatically determining process values ​​related to the defect situation based on the defect data set for each recognized defect situation, wherein the steps are performed by a defect analysis system (144), and in order to automatically determine the cause of the defect for the defect situation and / or automatically determine process values ​​related to the defect situation, one or more process values ​​are linked to the following link criteria: association of process values ​​with the same part of the industrial process plant (101) as where the defect situation occurred; linking process values ​​to historical defect situations based on active selection by a user; linking process values ​​from a message system that is pre-determined to be related to the defect situation; automatically linking to the defect situation based on one or more of the active selections of process values ​​by a user, automatically determining the cause of the defect for the defect situation and / or the defect situation A method characterized by automatically determining the relevant process values, additional defect causes and / or process values ​​being proposed, wherein the proposal is made based on one or more of the following proposal criteria: process relevance of the process values; the location of the process values ​​in the industrial process plant (101), or the location of the sensor determining the process values; the amount by which the process values ​​deviate from a defined process window and / or normal conditions; the prioritization of historical process values ​​in historical defect situations; and physical dependencies of the process values. Claim 2 delete Claim 3 In claim 1, a method for automatically determining the cause of a defect in the defect situation and / or the process values ​​associated with the defect situation, wherein the automatic prioritization of the process values ​​linked to the defect situation is performed automatically based on one or more of the following prioritization criteria: process relevance of the process values; the location of the process value in the industrial process plant (101), or the location of the sensor determining the process value; the amount by which the process value deviates from a defined process window and / or normal conditions; prioritization of historical process values ​​in historical defect situations; and prioritization of process values ​​from the cause of the defect and / or message system (138). Claim 4 delete Claim 5 In claim 1, a method in which historical defect situations are determined from a defect database (136) using one or more of the following similarity criteria: defect classification of said historical defect situation; historical defect situation of the same or similar plant part; process values ​​of said historical defect situation that are the same or similar to the process values ​​of said recognized defect situation. Claim 6 In claim 1, a method in which historical process values ​​identical or similar to the process values ​​of the recognized defect situation are determined from the process database (134). Claim 7 In claim 6, a method in which the determined history process values ​​are characterized as belonging to a history defect situation. Claim 8 In claim 1, for a recognized defect situation, the defect situation data set is stored in a defect database (136). Claim 9 A method according to claim 8, wherein each defect identification data set comprises one or more of the following defect situation data: defect classification of said defect situation; process values ​​linked to said defect situation based on pre-linking from a message system; information regarding the time when each defect situation occurred; information regarding the period during which each defect situation occurred; information regarding the location where each defect situation occurred; an alarm related to said defect situation; and a status message. Claim 10 In claim 8, the defect situation data set of each defect situation includes defect identification data for clear identification of the recognized defect situation. Claim 11 In claim 8, a method in which documentation data and defect removal data are stored in the defect situation data set of each defect situation. Claim 12 In claim 8, the process values ​​are stored during the operation of the industrial process plant (101) and synchronized with the recognized fault situation. Claim 13 A method according to claim 8, wherein process values ​​are provided with a time stamp, and the process values ​​are configured to be clearly associated with a predetermined point in time by the time stamp. Claim 14 A defect analysis system (144) for defect analysis in an industrial process plant (101), which takes the form for the purpose of performing a defect analysis method in an industrial process plant (101) according to any one of claims 1, 3 and 5 to 13, and is also configured for this purpose. Claim 15 An industrial control system (100) including a defect analysis system (144) according to claim 14.

Images