CN-122009431-A - Defect fault positioning method for ship navigation detection system

CN122009431ACN 122009431 ACN122009431 ACN 122009431ACN-122009431-A

Abstract

The invention relates to the technical field of ship electronics and automation, in particular to a defect fault positioning method for a ship navigation detection system, which comprises the following steps of obtaining execution data of each code module in the ship navigation detection system under a test case; extracting multidimensional fault characteristics of each code module, constructing a comprehensive feature vector of each code module, establishing an initial KAN confidence assessment model, training the initial KAN confidence assessment model by adopting a mixed sample data enhancement strategy, inputting the comprehensive feature vector into the trained KAN confidence assessment model, outputting fault confidence of each code module, sequencing each code module and determining the fault location of the defect. The invention can effectively fit the complex nonlinear fault mode of the codes in the ship navigation detection system, quickly and accurately position the defect faults in the ship navigation detection system, shorten the fault checking time and improve the reliability and development efficiency of the ship navigation detection system.

Inventors

GUO SHIKAI
Jiang Shuirou
Ni Anxiang

Assignees

大连海事大学

Dates

Publication Date: 20260512
Application Date: 20260407

Claims (8)

1. The defect fault positioning method for the ship navigation detection system is characterized by comprising the following steps of: Acquiring execution data of each code module in a ship navigation detection system under test cases, wherein the test cases comprise passing cases and failure cases; based on the execution data, extracting the multidimensional fault characteristics of each code module, splicing the multidimensional fault characteristics, and constructing a comprehensive feature vector of each code module; establishing an initial KAN confidence assessment model, wherein the initial KAN confidence assessment model adopts a learnable spline parameter as an activation function; Training the initial KAN confidence assessment model by adopting a mixed sample data enhancement strategy to obtain a trained KAN confidence assessment model; and inputting the comprehensive feature vector into the trained KAN confidence evaluation model, outputting the fault confidence of each code module, sequencing the code modules based on the fault confidence, and determining at least one code module which is sequenced to the front as a defect fault position.
2. The method for locating defects of a ship navigation detection system according to claim 1, wherein the step of obtaining execution data of each code module in the ship navigation detection system under test cases comprises constructing a test case set covering a typical navigation scene based on a ship navigation simulation environment, and running the test case set to record execution coverage of each code module.
3. The defect fault positioning method for the ship navigation detection system according to claim 2, wherein the ship navigation simulation environment comprises a normal navigation scene and a fault injection scene, the test cases comprise a passing case, a failing case and an auxiliary case, and the code module comprises a radar echo processing module, an AIS data decoding module, a Kalman filtering fusion module, an autonomous collision avoidance decision module and a navigation data display module.
4. A method of fault localization for a marine vessel voyage detection system according to claim 1, wherein the step of extracting the multi-dimensional fault signature of each code module comprises: Extracting suspicious characteristics based on spectrum analysis, wherein the suspicious characteristics based on spectrum analysis are used for representing statistical correlation between coverage frequency of a code module in a failure use case and a test result; Extracting suspicious characteristics based on mutation analysis, wherein the suspicious characteristics based on mutation analysis are used for representing the detection capability of a test case set on variants after the code modules are mutated by a mutation tool, and the suspicious characteristics based on mutation analysis are divided into a plurality of subsets according to the types of test results; extracting static characteristics based on code complexity, wherein the static characteristics based on the code complexity comprise ring complexity, nesting depth, holsteeld length and Holsteeld difficulty; and extracting a characteristic based on text similarity, wherein the characteristic based on text similarity is used for representing the text similarity between the descriptive text of the failed use case and the identification information of the code module.
5. The fault location method for the ship navigation detection system according to claim 4, wherein the suspicious characteristics based on the frequency spectrum analysis are obtained through calculation through a plurality of frequency spectrum formulas, the suspicious characteristics based on the mutation analysis are obtained through generating a plurality of variants on a code module and counting the detection capability of a test case set on the variants, and the characteristics based on the text similarity are obtained through word segmentation processing and similarity calculation on descriptive text of a failure case and module names, function names, signal names and notes of the code module.
6. The method for fault localization of a ship navigation detection system according to claim 4, wherein the initial KAN confidence assessment model comprises a plurality of hidden layers, each hidden layer is provided with a plurality of nodes, each node is mapped by a plurality of piecewise linear basis functions, spline coefficients of the basis functions are updated along with network training as learnable parameters, the initial KAN confidence assessment model adopts a hierarchical feature fusion strategy to fuse a plurality of subsets of suspicious features based on mutation analysis, and integrates the suspicious features based on spectrum analysis, the static features based on code complexity and the features based on text similarity hierarchically.
7. The defect fault locating method for the ship navigation detection system according to claim 1, wherein the mixed sample data enhancement strategy comprises the steps of sampling mixed weights from beta distribution, randomly selecting two real fault samples to mix, generating virtual training samples, wherein feature vectors of the virtual training samples are weighted averages of feature vectors of the two real fault samples, labels are soft labels corresponding to fault types, determining loss weights of the weighted cross entropy loss functions based on the mixed weights during training, and optimizing network parameters by adopting a gradient descent algorithm to minimize classification loss and regularization items.
8. The defect fault locating method for ship navigation detection system according to claim 1, wherein after determining at least one code module in the top order as the defect fault location, the defect fault locating result is evaluated, and the top N recall, the average ranking and the top fault ranking are used as evaluation indexes.

Description

Defect fault positioning method for ship navigation detection system Technical Field The invention relates to the technical field of ship electronics and automation, in particular to a defect fault positioning method for a ship navigation detection system. Background The ship navigation detection system is core equipment for guaranteeing the navigation safety of the ship, and provides navigation decision support for crews by collecting information such as ship position, course, speed, depth of water, surrounding ship dynamics and the like. Modern ship navigation detection systems are usually realized based on FPGA (field programmable gate array) or embedded processors, and software functions cover key tasks such as data fusion, target tracking, route planning, collision early warning and the like, and have the characteristics of high software complexity and severe reliability requirements. The FPGA is widely applied to functional modules with higher real-time requirements such as radar signal processing, multi-sensor data fusion, high-speed communication interfaces and the like in a ship navigation detection system due to the parallel processing capability and low delay characteristic. In the process of developing FPGA embedded software of a ship navigation detection system, development is generally performed by using a hardware description language, or modeling is performed by using a model-based design tool and a code is automatically generated. Because the system has complex functions and various interfaces, various defects are easily introduced in the code writing and code generating stages, including sensor data analysis errors, state machine jump anomalies, timing competition problems and alarm logic timing errors. These defects may cause abnormal display of navigation data, warning of missing/false alarms, and even cause navigation accidents when severe. Therefore, in the research and development process of the ship navigation detection system, the defects in the FPGA code are rapidly and accurately positioned, and the method has important significance for guaranteeing the ship navigation safety, shortening the system research and development period and reducing the offshore test cost. At present, in the field of FPGA embedded software testing for critical tasks, remarkable progress has been made in the aspect of positioning defects of embedded software, but when the FPGA embedded software is applied to embedded codes realized based on FPGA in a ship navigation detection system, the following technical defects still exist: Firstly, in the aspect of complex fault mode identification, the existing defect positioning method is mainly based on traditional spectrum analysis and statistical correlation calculation, and has insufficient capability of capturing complex fault characteristics specific to FPGA codes in a ship navigation detection system. Faults in FPGA code often manifest themselves as state machine jump anomalies, timing contention problems, cross-clock domain synchronization errors, and fusion anomalies for multiple parallel data streams. These faults have high nonlinearity and coupling, and conventional methods are difficult to fit effectively. Secondly, in terms of generalization capability under the condition of small samples, the FPGA code of the ship navigation detection system is extremely rare in the research and development test process of real high-quality fault samples. The existing defect positioning method lacks a targeted data enhancement mechanism, the positioning accuracy of the model is obviously reduced when the model faces to accidental faults or unseen navigation scenes, and the severe requirements of a ship navigation detection system on high reliability cannot be met. Therefore, a fault positioning method of the FPGA embedded code in a ship navigation detection system is highly required, and can effectively identify a complex fault mode in the FPGA code and maintain good generalization capability under a small sample condition. Disclosure of Invention In order to solve the problems of insufficient complex fault mode identification capability and poor generalization capability under the condition of small samples in the FPGA embedded code fault positioning of a ship navigation detection system in the prior art, the invention provides a defect fault positioning method for the ship navigation detection system, which comprises the following steps: s1, acquiring execution data of each code module in a ship navigation detection system under test cases, wherein the test cases comprise passing cases and failure cases. S11, constructing a test case set covering a typical sailing scene based on the ship sailing simulation environment. And S12, running the test case set and recording the execution coverage condition of each code module. The ship navigation simulation environment comprises a normal navigation scene and a fault injection scene, the test cases comprise a passing case,