35 results about "Causal analysis" patented technology

The present invention is directed to computer-based methods and system to perform root-cause analysis on an industrial process. The methods and system load process data for an industrial process from a historian database and build a hybrid first-principles and inferential model. The methods and system then executes the hybrid model to generate KPIs for the industrial process using the loaded process variables. The methods and system then selects a subset of the KPIs to represent an event occurring in the industrial process, and divides the data for the subset into multiple subset of time series. The system and methods select time intervals from the time series based on the data variability in the selected time intervals and perform a cross-correlation between the loaded process variables and the selected time interval, resulting in a cross-correlation score for each loaded process variable. The methods and system then select precursor candidates from the loaded process variables based on the cross-correlation scores and execute a parametric model for performing quantitative analysis of the selected precursor candidates, resulting in a strength of correlation score for each precursor candidate. The methods and system select root-cause variables from the selected precursor candidates based on the strength of correlation scores for analyzing the root-cause of the event.

The invention discloses a multivariable clustering and fusion time series combination prediction method; aiming at solving the problems that an existing neural network model does not have a specific learning mechanism and cannot fully mine data structure feature information, from the multivariable directed coupling perspective, and in combination with the advantages of a graph convolutional neuralnetwork and a long-term and short-term memory network, the invention provides a multivariable clustering and fusion time series combination prediction method. The method comprises the following steps: firstly, exploring a causal transfer relationship between variables based on coupled Granger causal measure analysis; secondly, establishing a directed weighted network according to a variable causality analysis result, extracting node and edge weight characteristics of the directed weighted network, and embedding the weight of a target variable into a graph convolutional neural network for training to realize accurate classification of monitoring variables; finally, taking the non-target monitoring variable time series contained in the community where the target monitoring variable is located as input, and predicting the target monitoring variable based on the long-term and short-term memory neural network. The method is applied to verification of a compressor unit monitoring sequence in a chemical production system, and results show that the method is superior to a traditional node classification method in the aspects of prediction accuracy and calculation complexity, and the proposed method can also maintain high prediction capability in an abnormal state of the system.

The invention discloses an analysis method for a causal network of multi-dimensional time-sequence physiological signals. The method comprises the following steps: S1. decomposing physiological signals, u1, u2, ..., um to be analyzed by using a NA-MEMD algorithm; S2. carrying out causal analysis on two different physiological signals ui and uj, so as to obtain a causal relationship between the twosignals, wherein i= 1, 2, ..., m, and j= 1, 2, ..., m; and S3. repeatedly executing the step S2 on any two signals of u1, u2, ..., um until causal relationships between every two signals of u1, u2, ..., um are obtained, thereby forming the causal network. According to the method, the causal network of the physiological signals can be effectively analyzed, and convenience is provided for application of the physiological signals.

The invention discloses a pathogenic microorganism metagenome-based inter-flora Granger causality analysis method, and the method comprises the following steps: performing clustering analysis on acquired pathogenic microorganism metagenome data, dividing each strain, and calculating the species abundance of each strain at a corresponding disease development time node; and then establishing a timesequence of each group of variables for each time node and correcting, and finally performing Granger causality analysis on the corrected data to find out main pathogenic bacteria. Various bacteria are identified by utilizing a high-throughput sequencing platform, and then the causal action relationship among the bacteria is found out by combining a Granger causal analysis method, so that a rapididentification method is provided for finding out one or more pathogenic bacteria from a pathogenic microorganism sample, and clinicians are helped to carry out analysis and judgment more rapidly to realize clinical diagnosis.

The embodiment of the invention belongs to the field of artificial intelligence, and relates to a system strategy prediction method based on Granger causality, comprising: collecting related parameters and industrial data, the related parameters comprising at least one module variable of at least one system module, and the industrial data being historical data of each module variable; based on the historical data, performing causal analysis on each module variable by adopting a Granger causal relationship analysis mode based on an attention mechanism to obtain Granger causal strength between the historical data of each module variable and the current data; constructing a graph data structure to describe a system topological structure of the industrial system; and inputting the constructed graph data structure into the prediction model to obtain a prediction result of the system strategy. By adopting the method, the model prediction accuracy is improved.

The invention discloses a fatigue state causal network method based on multi-source data information. The fatigue state causal network method comprises the following steps that 1, breathing and heartrate data related to human fatigue state are collected; 2, according to physiological data, an extrapolation fitting method is used for removing outliers; 3, two of the physiological data are selectedat random, and whether a correlational relationship exists between the two data or not is tested; 4, for the two physiological data with the correlational relationship, a granger causality analysis method is used for testing whether a causal relationship exists between the two physiological data or not; 5, two sets of the physiological data are selected in a circular manner, and the granger causality analysis method in the step is utilized; and 6, after all physiological data are traversed, a causal relationship network between variables is established. The fatigue state causal network methodhas the advantages of high accuracy, small calculation amount and high compatibility and can be used for the causal analysis and detection of fatigue status of workers on vehicles such as automobiles, ships and aircraft, and the accuracy of causal analysis of fatigue status based on multi-source data fusion is improved, so that the application value is high.

The invention provides a power grid saturation load prediction method and apparatus, and a terminal device. The method comprises the steps of obtaining a current value of an influence factor of a to-be-predicted power grid based on a preset Granger causality analysis result; inputting the current value of the influence factor of the to-be-predicted power grid into a preset Gaussian process regression model, and determining a load prediction value of the to-be-predicted power grid; determining a future planning value of an influence factor of the to-be-predicted power grid based on the load prediction value; and inputting the future planning values of the influence factors of the to-be-predicted power grid into a preset Gaussian process regression model, and determining a saturation load prediction value of the to-be-predicted power grid. According to the power grid saturation load prediction method and apparatus, and the terminal device provided by the invention, the Granger causality analysis is introduced to screen the external factors influencing the power grid load, so that unnecessary external factors are reduced, the problem of pseudo regression generated in the power grid saturation load prediction process is solved. And the accuracy of power grid saturation load prediction is improved.

The invention relates to an engineering safety quality accident tracing method and device based on a block chain, and the method comprises the steps: responding to an engineering safety quality accident tracing request, displaying a causal analysis graph and each preset analysis dimension, and enabling the causal analysis graph to have a plurality of preset analysis dimensions, at least comprising any two or more of people, materials, systems and environments; when an analysis request for one analysis dimension is obtained, loading factor information of the corresponding analysis dimension from a corresponding block of the engineering supervision block chain, wherein the analysis request is triggered based on a causal analysis graph; and obtaining an occurrence reason of the engineering safety quality accident determined based on the factor information. The method can improve the tracing efficiency.

The embodiment of the invention discloses a data display method based on an index tree and terminal equipment. A specific embodiment of the method comprises the following steps: receiving a target index input by a user; determining a calculation rule index; determining an analysis method; determining a candidate data sequence based on the target index, the calculation rule index and the analysis method; and determining target data based on the candidate data sequence. According to the implementation mode, the invention includes building indexes of data into an index tree model set; searching a target index tree model according to the index tree model set; and performing causal analysis, fruit cause analysis and bidirectional analysis on the target index tree model by using the determined analysis method. The accuracy of determining the target data can be improved, and the target equipment is controlled to improve the action completion level.

The invention discloses an industrial control system multi-loop oscillation causality analysis method based on an improved CCM. The method comprises the following steps: calculating mutual informationbetween every two oscillation circuit process output signals to be analyzed to perform feature selection, a loop signal pair with a relatively high correlation degree is reserved; removing noise andperiodic terms by using EMD and DFA methods; calculating a cross mapping index of the reconstructed sub-signal pair under different time delays by using a CCM method; preliminarily judging the authenticity of the causal relationship according to the positive and negative of the optimal time delay, then the cross mapping indexes of each target sub-signal pair corresponding to the optimal time delayunder different sample lengths are calculated, judging whether convergence occurs according to the convergence threshold, and finally obtaining the final causal relationship network of the oscillation circuit, so that the oscillation propagation path is determined and the oscillation source is positioned. According to the method, the oscillation propagation path in the industrial control system can be rapidly and accurately diagnosed, the oscillation source can be positioned, and a new thought is provided for fault diagnosis of the oscillation circuit of the industrial control system by usinga causal analysis method.

The invention relates to an artificial intelligence technology, and discloses a causal interpretation method based on image classification, which comprises the following steps: carrying out initial partitioning on a to-be-analyzed image based on superpixels and a linear iterative clustering algorithm to obtain a plurality of partitioned images; calculating responsibility degrees of the plurality of partition images respectively, and performing secondary partition on the partition images of which the responsibility degrees are greater than a responsibility degree threshold value to obtain a plurality of secondary partition images; the responsibility degrees of the secondary partition images are calculated respectively, and when the number of pixel points of the secondary partition images is smaller than or equal to a pixel point threshold value, or the responsibility degrees of the secondary partition images are equal, the secondary partition images serve as a standard attribution map; and performing causal analysis on the standard attribution graph to obtain a causal analysis result. In addition, the invention also relates to a block chain technology, and the partition image can be stored in a node of a block chain. The invention further provides a causal interpretation device based on image classification, electronic equipment and a storage medium. The accuracy of causal interpretation based on image classification can be improved.

The invention discloses an end-to-end brain causal network construction method based on a graph neural network, and belongs to the field of electroencephalogram information processing. According to the method, a multilayer perceptron with adjacent k layers of feature fusion is designed for multi-dimensional feature extraction, and a drawing neural network is further designed for direct mining of the brain causal relationship. Then, a multivariate sequence with real electroencephalogram signal characteristics and causal supervision information of the multivariate sequence are obtained through a vector autoregression model, and a neural network model is trained through a supervision method; and based on the trained neural network model, mining of the causal relationship of the electroencephalogram data and construction of the causal network can be realized. Compared with a representative method of a traditional method, Granger causal analysis comparative research proves that the method has remarkable advantages in the aspect of capturing the causal network topological structure and causal relationship strength under the condition of low signal-to-noise ratio. According to the method, a new perspective is provided for breaking through traditional model-driven hypothesis constraints and directly mining a deep brain causal network mechanism in a data-driven mode.

The invention relates to a causal analysis method and system for proving evidence based on prompts, and belongs to the field of natural language processing. The method mainly comprises the following steps of: (1) formalizing a structural causal model of an interaction relationship of each variable in a proving process based on a prompt; the structural causal model totally comprises 11 key variables, and describes causal relationships in four key processes of model pre-training, prompt selection, natural linguistic test set generation and performance evaluation. And (2) based on a structural causal model, identifying a real causal relationship expected to be evaluated and back door paths of three confused evaluation results, and analyzing three deviations caused by the three back door paths: prompt preference deviation, instance natural linguistic deviation and sampling difference deviation. And (3) a causal intervention method for eliminating probing process deviation based on a backdoor criterion. The method can effectively identify, understand and eliminate the deviation in the probing process, and obtains a stable, accurate and reliable evaluation result.

The invention discloses an intelligent logic analysis system. The intelligent logic analysis system comprises a cause and effect definition set, a cause and effect analysis algorithm set and a data set, the cause and effect definition set is a cause and effect classification set with determined meanings, any classification can be a cause, a cause and an effect, and each classification contains individual objects; wherein the cause and effect analysis algorithm set is an algorithm set for analyzing the cause and effect objects in the cause and effect set; wherein the data set is composed of different data required by a causal analysis algorithm. The method has the beneficial effects that algorithm three-dimensional construction of different results is obtained by implementing different analysis algorithms on various data, so that each algorithm corresponding to the data can interact in a three-dimensional space.

The invention relates to a non-linear extended Granger causality analysis method with zero time lag, which belongs to the technical field of data mining. The traditional Granger causal model is based on the VAR model of the lagged term, but ignoring the zero time lag will greatly change the model coefficients of the lagged term, leading to wrong causal identification. Secondly, the traditional Granger causal model can only be applied to the causal identification of linear systems, and may produce wrong causal identifications for nonlinear systems. Based on the above analysis, the present invention first expands the traditional VAR model, sets the lag order of the variables, uses the Gaussian kernel function to nonlinearly map the original data, and then establishes a structural VAR model containing zero lag items. Finally, according to the structure Granger causality identification is carried out on the residuals of the VAR model to realize the zero-delay causality analysis of nonlinear systems such as pollution and meteorology. The invention can overcome the deficiency of the traditional Granger causality model, and realize the extended Granger causality analysis of the nonlinear system.

The invention discloses a sparse causal analysis method of industrial control multi-loop oscillation behaviors. The method includes: collecting the process output time sequence signal of each to-be-analyzed control loop; performing cosine-removing trend preprocessing on each process output time sequence signal to obtain a corresponding sub-signal; building a sparse multi-variable vector autoregression model for the sub-signal of each control loop; performing Granger causality analysis on the basis of the sparse multi-variable vector autoregression models to obtain the Granger causality index between optional two control loops; performing statistical significance judgement on the Granger causality index, and determining that the causal relationship between the corresponding control loops indicated by the Granger causality index is established under the corresponding significance level if the result is significant. By the sparse causal analysis method, a multi-oscillation-loop oscillation source can be positioned, an oscillation propagation path can be detected, and effective reference can be provided for the diagnosing and checking of oscillation faults.

The invention discloses a subsystem fluctuation signal analysis method based on causal analysis, and the method comprises the steps: carrying out the decomposition of a multi-dimensional signal through employing an MEMD method, obtaining the Imf of each layer of all signals, carrying out the grouping of each layer Imf of each signal based on a normalized correlation coefficient, correcting the group based on a sparse index, and carrying out the analysis of a fluctuation signal. A harmonic detection method is adopted to search for fundamental waves and harmonic waves Imf which are in an integral multiple relation in the groups, signals with the fundamental waves and the harmonic waves at the same time are selected to serve as the result of grouping harmonic detection feature selection, denoising and periodic term removing processing is conducted on fluctuation signals subjected to feature selection, the remaining part is reconstructed, and corresponding target sub-signals are obtained; and calculating cross mapping indexes of the reconstructed sub-signal pairs under different time lag by using an extended CCM method, judging whether to converge according to a convergence threshold, obtaining a causal relationship of each signal pair, further obtaining a fluctuation propagation path, and positioning a fluctuation source. According to the method, the fluctuation propagation path in the industrial control system can be quickly and accurately diagnosed, and the fluctuation source can be positioned.

The invention discloses a redundant inertial measurement unit fault detection method based on Granger causal analysis. The method comprises the following steps: (1) acquiring inertial measurement unit nominal value data and multi-redundant inertial measurement unit measured value data; (2) constructing a TVAR model by using an inertial measurement unit nominal value, and performing model parameter identification and model reconstruction; (3) executing the following processing on each inertial measurement unit in the multiple redundant inertial measurement units: (a) constructing a TVAR model according to the measured value of the current inertial measurement unit, and performing model parameter identification and model reconstruction; (b) constructing an ectogenic input time-varying autoregressive representation model by taking the nominal value of the inertial measurement unit as output and the measured value of the current inertial measurement unit as input, and performing model parameter identification and model reconstruction; (c) taking the measured value of the current inertial measurement unit as output and the nominal value of the inertial measurement unit as input to construct a time-varying autoregressive representation model of exogenous input, and performing model parameter identification and model reconstruction; (d) calculating an interactive Granger causality index and an error of the interactive Granger causality index between each inertial measurement unit and the inertial measurement unit nominal value; and (4) setting an evaluation principle, and evaluating whether each inertial measurement unit has a fault or not.