Transition state aero-engine gas path potential multiple fault fusion diagnosis method

Abstract

The application discloses a transition state aero-engine gas path potential multiple fault fusion diagnosis method, and belongs to the field of aero-engine fault diagnosis. The transition state aero-engine gas path potential multiple fault fusion diagnosis method comprises the following steps: researching aero-engine gas path transition state data cooperative noise reduction and multi-dimensional representation to obtain multi-dimensional gas path transition state sample data; constructing an aero-engine gas path transition state potential multiple fault fusion diagnosis model, training the diagnosis model by using the multi-dimensional gas path transition state sample data, and outputting a fault diagnosis result by using the trained aero-engine gas path transition state potential multiple fault fusion diagnosis model; explaining the aero-engine gas path transition state potential multiple fault fusion diagnosis model and revealing a fault mechanism; and realizing accurate mapping from limited gas path monitoring information to potential multiple faults and mechanism revealing, thereby providing theoretical and application support for aero-engine explainable intelligent operation and maintenance.

Description

Technical Field

[0001] This invention belongs to the field of aero-engine fault diagnosis, specifically relating to a method for fusion diagnosis of potential multiple faults in the aero-engine gas path under transient conditions. Background Technology

[0002] During the steady-state operation of aero-engines, indirect fault diagnosis can be achieved by mapping the deviation of monitoring parameters to the performance changes of gas path components. However, in the transitional state where the steady state rapidly changes, engine speed, power, and aerodynamic and thermodynamic parameters all change drastically and nonlinearly over time. The operating lines of components are closer to the fault boundaries of surge, stall, and in-flight shutdown, making it difficult to accurately characterize the transitional state condition with limited monitoring information. At the same time, the impact of instantaneous measurement interference is greater in the transitional state, the fault mechanism is more complex, and multiple faults are more likely to occur, leading to functional failure. While domestic and international research has accumulated rich achievements in aero-engine gas path monitoring data acquisition and preprocessing, data-driven aero-engine gas path fault diagnosis, and improvement of the interpretability of fault diagnosis models, most current research focuses on gas path fault diagnosis under steady-state conditions. Further research is needed on the construction of gas path monitoring samples, diagnosis of potential multiple faults, and interpretable fault cognition under the transitional state where operating conditions change rapidly and monitoring parameters fluctuate drastically. Therefore, this paper proposes a fusion diagnosis method for potential multiple faults in aero-engine gas paths under transitional conditions. Summary of the Invention

[0003] To address the shortcomings of existing technologies, the present invention aims to provide a method for fusion diagnosis of potential multiple faults in the gas path of aero-engines under transitional conditions, thus solving the problems in existing technologies.

[0004] The objective of this invention can be achieved through the following technical solutions:

[0005] A fusion diagnostic method for potential multiple faults in the gas path of an aero-engine under transient conditions includes the following steps:

[0006] This study investigates collaborative noise reduction and multi-dimensional characterization of transition state data in aero-engine air paths in order to obtain multi-dimensional transition state sample data in air paths.

[0007] A fusion diagnostic model for potential multiple faults in the transition state of the aero-engine's gas path is constructed, and the diagnostic model is trained using multi-dimensional gas path transition state sample data. The fault diagnosis results are then output using the trained fusion diagnostic model for potential multiple faults in the transition state of the aero-engine's gas path.

[0008] The fusion diagnostic model for potential multiple faults in the transition state of the aero-engine gas path is explained, and the fault mechanism is revealed.

[0009] Furthermore, the process of obtaining multi-dimensional gas path transition state sample data is as follows:

[0010] 1) Input multi-parameter time-series samples with added random noise into a multi-head self-attention Seq2Seq model, encode input samples of different sizes using bidirectional gated recurrent units, and achieve feature focusing through a multi-head self-attention mechanism;

[0011] 2) The reconstructed sample is obtained through the decoding operation of the multi-head attention bidirectional gated recurrent unit. The error between the original noiseless sample and the reconstructed sample is used as the loss function to back-optimize the multi-head self-attention sequence to sequence model, and the collaborative noise reduction result of multi-monitoring parameter variable-scale time series sample is obtained.

[0012] 3) Using a nonlinear pseudo-nearest neighbor algorithm and average mutual information method, appropriate embedding dimensions and delay times are selected to obtain delay vectors and reconstruct high-dimensional phase space to construct multi-dimensional gas path transition state sample data.

[0013] Furthermore, the aero-engine gas path transition state potential multiple fault fusion diagnostic model includes: a potential fault reasoning unit, a multiple fault decoupling unit, and a deep meta-transfer learning diagnostic unit;

[0014] Specifically, the fault features obtained from the multi-dimensional gas path transition state sample data are processed by a feature extractor and then input into a sample weight learner, a potential fault inference unit, and a deep meta-transfer learning diagnostic unit based on a multi-fault decoupling unit. The comprehensive loss function is a weighted sum of the inference results, diagnostic results, and sample weights. The optimal fusion diagnostic model is obtained by iteratively learning the model parameters and sample weights through backpropagation of the comprehensive loss function.

[0015] Furthermore, the process by which the potential fault reasoning unit infers potential faults is as follows:

[0016] A probability map consisting of known gas path monitoring samples, unknown confounding factors, transitional operating conditions, and potential faults is constructed. The inference network is encoded into a multivariate Gaussian distribution and then decoded to obtain a model network. Potential faults are inferred based on the conditional probability distribution.

[0017] Furthermore, the multi-fault decoupling unit inputs the fault features obtained by processing the multi-dimensional gas path transition state sample data through the feature extractor into a smooth Wigner-Ville distribution for non-stationary signal feature extraction, and then uses random Fourier features to project it into a high-dimensional space to eliminate the correlation of multi-fault coupling features and learn fault weights.

[0018] Furthermore, the deep meta-transfer learning diagnostic unit selects a deep residual convolutional neural network with unidirectional kernel movement as the learner, and combines it with a meta-learning method that employs gradient consistency and loss approximation strategies to fine-tune the meta-learner and the base learner. The meta-learner and the base learner are respectively used as the known working condition source domain model and the new working condition target domain model of the transfer learning model, and the deviation of the same hidden layer of the source domain model and the target domain model in extracting features for the same sample is calculated. The hidden layer feature deviation exceeding the threshold is added to the target domain loss function to optimize the meta-transfer learning model.

[0019] A fusion diagnostic system for potential multiple faults in the gas path of an aero-engine under transitional conditions, including:

[0020] Sample data acquisition module: Research on collaborative noise reduction and multi-dimensional characterization of aero-engine gas path transition state data to obtain multi-dimensional gas path transition state sample data;

[0021] Diagnostic model construction module: Constructs a fusion diagnostic model for potential multiple faults in the transition state of the aero-engine air path, trains the diagnostic model using multi-dimensional air path transition state sample data, and outputs fault diagnosis results using the trained fusion diagnostic model for potential multiple faults in the transition state of the aero-engine air path.

[0022] In addition, the diagnostic model interpretation module interprets the fusion diagnostic model of potential multiple faults in the transition state of the aero-engine air path and reveals the fault mechanism.

[0023] A computer storage medium storing a readable program, which, when the program is run, enables the aforementioned method for fusion diagnosis of potential multiple faults in the air circuit of an aero-engine under transitional conditions.

[0024] An electronic device includes: a processor, a memory, a communication interface, and a communication bus, wherein the processor, the memory, and the communication interface communicate with each other through the communication bus;

[0025] The memory is used to store at least one executable instruction, which causes the processor to perform the operation corresponding to the above-described method for fusion diagnosis of potential multiple faults in the aero-engine gas path under transitional conditions.

[0026] A computer program product includes computer instructions that instruct a computing device to perform operations corresponding to the above-described method for fusion diagnosis of potential multiple faults in the air path of an aero-engine under transitional conditions.

[0027] The beneficial effects of this invention are:

[0028] 1. This invention transforms aero-engine gas path monitoring data under low signal-to-noise ratio and poor information background into high-quality multi-dimensional samples that characterize the complex transition state process, thereby realizing the perception of gas path transition state information.

[0029] 2. This invention constructs a precise fusion diagnostic model for potential multiple faults in the gas path of aero-engines under transitional state by using counterfactual reasoning of potential causality and decoupling of multiple faults under strong correlation.

[0030] 3. This invention improves the reliability of gas path transition state fusion fault diagnosis results by using an interpretable attribution and graphical understanding of fault mechanisms through a fusion diagnostic model of potential multiple faults in the gas path transition state. Attached Figure Description

[0031] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0032] Figure 1 This is a flowchart of the fault diagnosis process of the present invention;

[0033] Figure 2 This is a flowchart of the transitional state data collaborative noise reduction and multi-dimensional characterization process of the present invention;

[0034] Figure 3 This is a framework diagram of the fault diagnosis model of the present invention. Detailed Implementation

[0035] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0036] Example 1

[0037] like Figure 1 As shown, the method for fusion diagnosis of potential multiple faults in the gas path of an aero-engine under transient state includes the following steps:

[0038] S1, to study collaborative noise reduction and multi-dimensional characterization of transition state data of aero-engine gas path, in order to obtain multi-dimensional transition state sample data of gas path;

[0039] To address the challenges of retaining fault information while reducing noise in transitional gas path data under low signal-to-noise ratio conditions, and the difficulty in characterizing the transitional system state with short-time data of limited gas path monitoring parameters, this invention studies the collaborative noise reduction and multi-dimensional characterization of aero-engine gas path transitional data, thereby obtaining high-quality multi-dimensional gas path transitional samples for subsequent fault feature extraction and diagnostic model training.

[0040] like Figure 2 As shown, the process of collaborative noise reduction of aero-engine gas path transition state data is as follows: time series data of finite gas path monitoring parameters with variable length are collected under the transition state, and after adding random perturbation, they are input into a multi-head self-attention Seq2Seq model to obtain reconstructed data. By calculating the error between the reconstructed data and the original data without interference and performing reverse optimization, collaborative noise reduction of variable time series data of multiple monitoring parameters is achieved.

[0041] Specifically, firstly, multi-parameter time-series samples with added random noise are input into a multi-head self-attention Seq2Seq model; then, bidirectional gated recurrent units are used to encode input samples of different sizes, and feature focusing is achieved through a multi-head self-attention mechanism; then, the reconstructed samples are obtained through decoding operations by the multi-head attention bidirectional gated recurrent units, and the error between the original noise-free samples and the reconstructed samples is used as a loss function to back-optimize the multi-head self-attention sequence to the sequence model, thus obtaining the collaborative noise reduction result of multi-monitoring parameter variable-scale time-series samples.

[0042] The process of dimensional characterization of aero-engine gas path transition state data is as follows: A nonlinear spurious nearest neighbor algorithm and average mutual information method are used to select appropriate embedding dimensions and delay times, obtaining a delay vector and reconstructing a high-dimensional phase space. This allows for the construction of multi-dimensional gas path transition state samples based on an understanding of the nonlinear dynamic characteristics of the gas path, used to characterize the complex time-varying patterns of gas path transition state monitoring data.

[0043] Specifically, after determining the embedding dimension and delay time, the monitoring data can be mapped to a high-dimensional space, so that at each time step, a sample of size (number of monitoring parameters × embedding dimension) can be obtained to characterize the nonlinear change characteristics of the gas path transition state.

[0044] S2, construct a fusion diagnostic model for potential multiple faults in the transition state of the aero-engine air path, and use the multi-dimensional transition state sample data of the air path obtained in S1 to train the diagnostic model, and use the trained fusion diagnostic model for potential multiple faults in the transition state of the aero-engine air path to output the fault diagnosis results.

[0045] like Figure 3As shown, the fusion diagnostic model for potential multiple faults in the transition state of an aero-engine's gas path includes: a potential fault inference unit, a multiple fault decoupling unit, and a deep meta-transfer learning diagnostic unit. A probability graph is constructed, consisting of known gas path monitoring samples, unknown confounding factors, transition state conditions, and potential faults. The inference network is encoded into a multivariate Gaussian distribution and then decoded to obtain the model network, thereby inferring potential faults based on the conditional probability distribution. Transition state fault features are input into a smoothed Wigner-Ville distribution for non-stationary signal feature extraction, and then projected into a high-dimensional space using stochastic Fourier features to eliminate the correlation of multiple fault coupling features and learn fault weights. Combining the potential fault inference, multiple fault decoupling, and deep meta-transfer learning diagnostic models, the fusion diagnostic result for potential multiple faults in the transition state is obtained through multi-model parameter optimization.

[0046] 1) Potential Fault Reasoning Unit

[0047] By using a causal variational autoencoder to analyze the impact of different fault types on the transient gas path state, we obtained the transient counterfactual potential fault inference results considering unknown factors.

[0048] The process of reasoning about potential faults is as follows:

[0049] A probability map consisting of known gas path monitoring samples, unknown confounding factors, transitional operating conditions, and potential faults is constructed. The inference network is encoded into a multivariate Gaussian distribution and then decoded to obtain a model network, thereby inferring potential faults based on the conditional probability distribution.

[0050] 2) Multiple Fault Decoupling Unit

[0051] The joint time-frequency analysis method is used to extract the non-stationary signal features of the transient fault characteristics and eliminate the correlation of multiple fault coupling features, thereby accurately distinguishing the multiple fault characteristics in the gas path transient condition.

[0052] The process of decoupling multiple faults is as follows:

[0053] The fault features obtained from the multi-dimensional gas path transition state sample data acquired in S1 are processed by the feature extractor and then input into a smooth Wigner-Ville distribution for non-stationary signal feature extraction. The random Fourier features are then used to project the fault features into a high-dimensional space to eliminate the correlation of multiple fault coupling features and to learn the fault weights.

[0054] 3) Deep Meta-Learning Diagnostic Unit: A deep residual convolutional neural network with unidirectional kernel movement is selected as the learner, combined with a meta-learning method employing gradient consistency and loss approximation strategies to fine-tune the meta-learner and base learner. The meta-learner and base learner are used as the known source domain model and the new target domain model for the transfer learning model, respectively. The bias in feature extraction for the same sample by the same hidden layer of the source domain model and the target domain model is calculated, and the hidden layer feature bias exceeding the threshold is added to the target domain loss function to optimize the meta-transfer learning model.

[0055] By combining potential fault reasoning, multiple fault decoupling, and deep meta-transfer learning diagnostic units, the transition state potential multiple fault fusion diagnostic results are obtained through multi-unit parameter optimization.

[0056] The steps for training the diagnostic model using the multi-dimensional gas path transition state sample data obtained in S1 are as follows:

[0057] Step 1: Calculate the weights of the input time-series fault samples using the sample weight learner. Combine the outputs of the potential fault inference unit and the deep meta-transfer learning diagnostic unit to calculate the inference error and diagnostic error. Then, weight these errors according to the sample weights to generate a comprehensive training error.

[0058] Step 2: Optimize the model using the backpropagation algorithm, including the parameters of the sample weight learner, the latent fault inference model, and the deep meta-transfer learning diagnostic model. The goal is to minimize the overall training error, thereby improving the model's ability to decouple fault modes and its inference accuracy.

[0059] Step 3: After completing the model training, the results of the potential fault reasoning model and the deep meta-transfer learning diagnostic model are integrated. Based on the diagnostic contribution of the sample weight allocation, the final diagnostic result is obtained, realizing accurate diagnosis of complex multi-faults and efficient model adaptability.

[0060] S3 interprets the fusion diagnostic model of potential multiple faults in the transition state of the aero-engine air path and reveals the fault mechanism;

[0061] To address the pain point of the "black box" nature of fault fusion diagnostic models and the challenge of understanding the model's decision-making process and fault mechanism, this invention studies the interpretation of the transitional state fault diagnosis model for aero-engine gas paths and reveals the fault mechanism, thereby improving the reliability of the transitional state fusion diagnostic results.

[0062] 1) Interpretation of the fault fusion diagnostic model

[0063] A multi-level game interaction method is used to solve the contribution of transitional features to the fault fusion diagnosis results, and the transitional feature saliency map is obtained by combining the transitional features and the unit contribution. A hierarchical class activation map is constructed to backpropagate the transitional feature saliency map to obtain the importance of each hidden layer in the feature extractor. Through multi-level feature map fusion, the input layer is reversed on the basis of unified modeling of feature visualization interpretation and expression ability.

[0064] 2) Revealing the transitional state fault mechanism of the gas path based on time-series knowledge graphs

[0065] After obtaining the gas path transition state fusion fault diagnosis results and the corresponding key monitoring information at the input end, it is necessary to sort out their logical relationships and form an understanding of the fault mechanism. Based on considering the complex temporal relationships of the transition state, this invention constructs a temporal knowledge graph that integrates existing monitoring information and fault diagnosis results into entities and embeds relationships, thereby dynamically and intuitively understanding the fault mechanism.

[0066] Based on a similar inventive concept, embodiments of the present invention also provide a computer storage medium storing a readable program that, when the program is run, can execute the above-described method for fusion diagnosis of potential multiple faults in the air path of an aero-engine under transitional conditions.

[0067] Based on a similar inventive concept, this invention provides an electronic device, including: a processor, a memory, a communication interface, and a communication bus, wherein the processor, the memory, and the communication interface communicate with each other through the communication bus;

[0068] The memory is used to store at least one executable instruction, which causes the processor to perform the operation corresponding to the above-described method for fusion diagnosis of potential multiple faults in the aero-engine gas path under transitional conditions.

[0069] Based on a similar inventive concept, embodiments of the present invention also provide a computer program product, including computer instructions, which instruct a computing device to perform the operations corresponding to the above-described method for fusion diagnosis of potential multiple faults in the air path of an aero-engine under transitional conditions.

[0070] Example 2

[0071] This embodiment discloses a fusion diagnostic system for potential multiple faults in the air path of an aero-engine under transitional conditions, specifically including:

[0072] Sample data acquisition module: Research on collaborative noise reduction and multi-dimensional characterization of aero-engine gas path transition state data to obtain multi-dimensional gas path transition state sample data;

[0073] Diagnostic model construction module: Constructs a fusion diagnostic model for potential multiple faults in the transition state of the aero-engine air path, trains the diagnostic model using multi-dimensional air path transition state sample data, and outputs fault diagnosis results using the trained fusion diagnostic model for potential multiple faults in the transition state of the aero-engine air path.

[0074] In addition, the diagnostic model interpretation module interprets the fusion diagnostic model of potential multiple faults in the transition state of the aero-engine air path and reveals the fault mechanism.

[0075] The methods of the present invention can be implemented in hardware, firmware, or as software or computer code that can be stored in a recording medium (such as a CD-ROM, RAM, floppy disk, hard disk, or magneto-optical disk), or as computer code originally stored on a remote recording medium or a non-transitory machine-readable medium and subsequently stored on a local recording medium, downloaded via a network. Thus, the methods described herein can be processed by software stored on a recording medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware (such as an ASIC or FPGA). It is understood that the computer, processor, microprocessor controller, or programmable hardware includes storage components (e.g., RAM, ROM, flash memory, etc.) capable of storing or receiving software or computer code that, when accessed and executed by the computer, processor, or hardware, implements the methods described herein. Furthermore, when a general-purpose computer accesses the code used to implement the methods shown herein, the execution of the code transforms the general-purpose computer into a dedicated computer for performing the methods shown herein.

[0076] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the claimed invention.

Claims

1. A method for fusion diagnosis of potential multiple faults in the gas path of an aero-engine under transient state, characterized in that, Includes the following steps: This study investigates collaborative noise reduction and multi-dimensional characterization of transition state data in aero-engine air paths in order to obtain multi-dimensional transition state sample data in air paths. A fusion diagnostic model for potential multiple faults in the transition state of the aero-engine's gas path is constructed, and the diagnostic model is trained using multi-dimensional gas path transition state sample data. The fault diagnosis results are then output using the trained fusion diagnostic model for potential multiple faults in the transition state of the aero-engine's gas path. An interpretation of the fusion diagnostic model for potential multiple faults in the transition state of aero-engine gas path is presented, and the fault mechanism is revealed. The process of obtaining multi-dimensional gas path transition state sample data is as follows: 1) Input multi-parameter time-series samples with added random noise into a multi-head self-attention Seq2Seq model, encode input samples of different sizes using bidirectional gated recurrent units, and achieve feature focusing through a multi-head self-attention mechanism; 2) The reconstructed sample is obtained through the decoding operation of the multi-head attention bidirectional gated recurrent unit. The error between the original noiseless sample and the reconstructed sample is used as the loss function to back-optimize the multi-head self-attention sequence to sequence model, and the collaborative noise reduction result of multi-monitoring parameter variable-scale time series sample is obtained. 3) Using a nonlinear pseudo-nearest neighbor algorithm and average mutual information method, appropriate embedding dimensions and delay times are selected to obtain delay vectors and reconstruct high-dimensional phase space to construct multi-dimensional gas path transition state sample data; The aero-engine gas path transition state potential multiple fault fusion diagnostic model includes: a potential fault reasoning unit, a multiple fault decoupling unit, and a deep meta-transfer learning diagnostic unit. Specifically, the fault features obtained from the multi-dimensional gas path transition state sample data are processed by the feature extractor and then input into the sample weight learner, the potential fault inference unit, and the deep meta-transfer learning diagnostic unit based on the multi-fault decoupling unit. The comprehensive loss function is the weighted sum of the inference result, the diagnostic result, and the sample weight. The optimal fusion diagnostic model is obtained by iteratively learning the model parameters and sample weight through backpropagation of the comprehensive loss function. The process by which the potential fault reasoning unit infers potential faults is as follows: A probability map consisting of known gas path monitoring samples, unknown confounding factors, transitional operating conditions, and potential faults is constructed. The inference network is encoded into a multivariate Gaussian distribution and then decoded to obtain the model network. Potential faults are inferred based on the conditional probability distribution. The multi-fault decoupling unit inputs the fault features obtained by processing the multi-dimensional gas path transition state sample data through the feature extractor into a smooth Wigner-Ville distribution for non-stationary signal feature extraction, and then uses random Fourier features to project it into a high-dimensional space to eliminate the correlation of multi-fault coupling features and learn fault weights. The deep meta-transfer learning diagnostic unit uses a deep residual convolutional neural network with unidirectional kernel movement as the learner, and combines it with a meta-learning method that employs gradient consistency and loss approximation strategies to fine-tune the meta-learner and the base learner. The meta-learner and the base learner are used as the known working condition source domain model and the new working condition target domain model of the transfer learning model, respectively. The deviation of the same hidden layer of the source domain model and the target domain model in extracting features for the same sample is calculated, and the hidden layer feature deviation exceeding the threshold is added to the target domain loss function to optimize the meta-transfer learning model.

2. A fusion diagnostic system for potential multiple faults in the gas path of an aero-engine under transition state, comprising the method described in claim 1, characterized in that, include: Sample data acquisition module: Research on collaborative noise reduction and multi-dimensional characterization of aero-engine gas path transition state data to obtain multi-dimensional gas path transition state sample data; Diagnostic model construction module: Constructs a fusion diagnostic model for potential multiple faults in the transition state of the aero-engine air path, trains the diagnostic model using multi-dimensional air path transition state sample data, and outputs fault diagnosis results using the trained fusion diagnostic model for potential multiple faults in the transition state of the aero-engine air path. In addition, the diagnostic model interpretation module interprets the fusion diagnostic model of potential multiple faults in the transition state of the aero-engine air path and reveals the fault mechanism.

3. A computer storage medium storing a readable program, characterized in that, When the program runs, it can perform the fusion diagnosis method for potential multiple faults in the air circuit of an aero-engine under the transition state as described in claim 1.

4. An electronic device, characterized in that, include: The processor, memory, communication interface, and communication bus are provided, wherein the processor, memory, and communication interface communicate with each other via the communication bus. The memory is used to store at least one executable instruction, which causes the processor to perform the operation corresponding to the fusion diagnosis method for potential multiple faults in the air path of an aero-engine under transitional state as described in claim 1.

5. A computer program product comprising computer instructions, characterized in that, The computer instructions instruct the computing device to perform the operations corresponding to the transitional state aero-engine gas path potential multiple fault fusion diagnosis method as described in claim 1.

Images