Aircraft manufacturing equipment abnormity diagnosis method based on federated learning

Abstract

The invention relates to the field of federated learning, in particular to an aircraft manufacturing equipment abnormity diagnosis method based on federated learning. According to the method, firstly, a federal learning training framework is adopted for training of an aircraft manufacturing equipment anomaly detection model, the privacy security of data is guaranteed, and the privacy security of the aircraft manufacturing equipment data is further enhanced. On the basis of guaranteeing the data privacy security, the model for detecting the abnormality of the aircraft manufacturing equipment is trained, and model training is carried out on the data of the aircraft manufacturing equipment under the condition that the data are not out of local equipment through a federal learning architecture, so that the privacy security of the data is guaranteed. According to the method, a contract theory incentive mechanism is adopted, reasonable reward is provided for the aircraft manufacturing equipment to participate in the federated learning training process, the aircraft manufacturing equipment is incentive to participate in the federated learning training, and the accuracy of the aircraft manufacturing equipment anomaly detection model is improved.

Description

Technical Field

[0001] The present invention relates to the field of federated learning, and in particular to an abnormality diagnosis method for aircraft manufacturing equipment based on federated learning. Background Art

[0002] In recent years, with the intensified market competition, order fluctuations and increased attention to the full life cycle cost control of aircraft, the aircraft manufacturing supply chain is undergoing a "small core, large collaboration" transformation. The main engine factory undertakes the main business of final assembly, plays the leading role of the "chain leader" of the aviation manufacturing industry chain supply chain, and transforms from directly controlling the production process to using the resources of external aviation parts manufacturing enterprises to complete product manufacturing. The participation of aviation parts manufacturing enterprises can greatly improve the efficiency of aviation parts manufacturing, but there may be equipment abnormalities in the process of aircraft manufacturing equipment producing parts. Traditional equipment abnormality solutions usually require relevant professional and technical personnel to solve. With the rapid development of "small core, large collaboration", there are more and more aviation parts manufacturing enterprises in the supply chain, but most small and medium-sized aviation parts manufacturing enterprises do not have highly professional technical personnel, and traditional equipment abnormality solutions are no longer applicable to the current situation. In view of equipment abnormalities in the process of aircraft manufacturing equipment producing parts, it is necessary to use the data of aircraft manufacturing equipment to train the model in order to diagnose the abnormal causes of the equipment. However, information protection between different enterprises, training based on a certain enterprise data is prone to data leakage. In order to provide an efficient and safe aircraft manufacturing equipment abnormality diagnosis method, a model training method is urgently needed under the premise of protecting data privacy security.

[0003] Federated learning is a distributed machine learning architecture that ensures data privacy and security. Under the federated learning framework, aircraft manufacturing equipment data can complete the training of federated learning model tasks without leaving the local device. Aircraft manufacturing equipment needs to consume certain computing resources to complete the training of the federated learning model during the training of the federated learning task. Therefore, the publisher of the federated learning task should provide a reward mechanism to pay the corresponding remuneration to the aircraft manufacturing equipment participating in the federated learning training. Due to the information asymmetry problem between the federated learning center server and the aircraft manufacturing equipment in the federated learning task, the federated learning center server cannot confirm the workload of the aircraft manufacturing equipment. The incentive method based on contract theory is an effective way to promote honest behavior, especially in the case of information asymmetry.

[0004] The incentive mechanism based on contract theory can motivate aircraft manufacturing equipment to complete federated learning training tasks according to their own training capabilities under the condition of information asymmetry. Under this incentive mechanism, aircraft manufacturing equipment can only obtain the highest benefits when choosing contracts that match their own training capabilities, so rational aircraft manufacturing equipment will choose the corresponding contracts according to the principle of maximizing benefits, so this incentive mechanism is self-revealing. The federated learning center server can judge the workload of aircraft manufacturing equipment based on the contracts selected by the aircraft manufacturing equipment, better train the federated learning model, and provide a more accurate abnormal diagnosis model for aircraft manufacturing equipment. Summary of the invention

[0005] In order to address the deficiencies in the prior art, the present invention provides an aircraft manufacturing equipment abnormality diagnosis method based on federated learning. Based on aircraft manufacturing equipment as the background, the present invention provides a distributed machine learning method for the aircraft manufacturing equipment abnormality diagnosis method on the basis of protecting data privacy, and at the same time solves the cooperation problem between the federated learning center server and the aircraft manufacturing equipment under the condition of information asymmetry.

[0006] The technical solution adopted by the present invention is:

[0007] A method for abnormal diagnosis of aircraft manufacturing equipment based on federated learning, such as Figure 1 As shown, the following steps are included:

[0008] Step 1: After receiving the task, the federated learning center server establishes a utility model of the federated learning center server and aircraft manufacturing equipment, satisfies individual rationality, incentive compatibility and monotonicity constraints, and constructs a contract model.

[0009] Step 2: The Federated Learning Center server signs a contract with the aircraft manufacturer;

[0010] After the Federated Learning Center server completes the construction of the optimal contract, it broadcasts the relevant tasks and the optimal contract to the aircraft manufacturing equipment. The aircraft manufacturer chooses to sign different types of optimal contracts based on the training capabilities of the equipment. The aircraft manufacturer that has signed the contract proceeds to step 3.

[0011] Step 3: The aircraft manufacturing equipment downloads the global model from the federated learning center server;

[0012] In the first round of training, the Federated Learning Center server will provide an initialization model, and the aircraft manufacturing equipment will download the initialization model for training. In subsequent rounds of training, the aircraft manufacturing equipment will download the global model updated in the previous round from the center server for training.

[0013] Step 4: The aircraft manufacturing equipment performs local training of the federated learning global model;

[0014] Step 4.1: The aircraft manufacturing equipment uses the equipment operation data generated by the local equipment to train the downloaded global model, and after the training is completed, execute step 4.2;

[0015] Step 4.2: The aircraft manufacturing equipment uploads the trained model to the federated learning center server.

[0016] Step 5: The federated learning center server evaluates the local model uploaded by the aircraft manufacturing equipment.

[0017] The federated learning central server will evaluate the local models uploaded by each aircraft manufacturing equipment participating in the federated learning training. If the model meets the acceptable requirements, the central server will receive the model and execute step 6. Otherwise, the model will be rejected and no corresponding compensation will be issued.

[0018] Step 6: The federated learning center server aggregates the local models uploaded by the aircraft manufacturing equipment according to the contract selection;

[0019] The federated learning center server will aggregate the federated learning model according to the selected contract for the local models uploaded by the aircraft manufacturing equipment that meets the requirements.

[0020] Step 7: The Federated Learning Center server records the corresponding remuneration of the aircraft manufacturer according to the contract;

[0021] After each round of federated learning training, the federated learning center server will update the remuneration payable to the aircraft manufacturer for participating in the federated learning global model aggregation according to the signed contract. When the phased federated learning training task is completed, the federated learning center server will issue the corresponding remuneration to the aircraft manufacturer.

[0022] Next, the federated learning center server and the aircraft manufacturing equipment conduct n (n>0) rounds of steps 3 to 7 of the federated learning training process, as shown in Figure 2 shown.

[0023] Step 8: Deploy the aircraft manufacturing equipment anomaly diagnosis model

[0024] The trained aircraft manufacturing equipment anomaly diagnosis model is deployed locally to the aircraft manufacturing equipment to detect equipment anomalies.

[0025] The specific process of step 1 includes the following steps:

[0026] Step 1.1: After receiving the training task, the federated learning center server constructs the utility model of the aircraft manufacturing equipment and the federated learning center server according to the private information characteristics of the aircraft manufacturing equipment, the training cost, and the model benefit of the federated learning center server.

[0027] Step 1.1.1: The federated learning center server builds the cost model C for training the federated learning task of aircraft manufacturing equipment of type i i .

[0028] Step 1.1.2: The federated learning center server builds the utility model U for training the federated learning task of aircraft manufacturing equipment of type i f , the utility model is the reward R gained from training the federated learning task for the aircraft manufacturing equipment i Subtract the cost C required to train the task i .

[0029] U f =R i -C i (1)

[0030] Step 1.1.3: The benefit of the local model uploaded by aircraft manufacturing equipment of type i for the federated learning task is Ψ i The utility U obtained by the federated learning central server from the models uploaded by the aircraft manufacturing equipment participating in the training s The profit of its global model minus the compensation paid to each aircraft manufacturing equipment.

[0031]

[0032] Where N represents the total number of aircraft manufacturing equipment participating in the federated learning task, I represents the total number of aircraft manufacturing equipment types, and ρ i represents the prior probability of aircraft manufacturing equipment of type i, ρ i ∈[0,1].

[0033] Step 1.2: The federated learning center server ensures that the contract model meets the basic constraints acceptable to aircraft manufacturing equipment based on individual rationality constraints, incentive compatibility constraints and monotonicity constraints, and constructs a feasible contract model.

[0034] Step 1.2.1: The utility of aircraft manufacturing equipment needs to satisfy the constraints of individual rationality. The aircraft manufacturing equipment of type i chooses contract φ i When non-negative utility is obtained.

[0035] U f =R i -C i ≥0 (3)

[0036] Step 1.2.2: Aircraft manufacturing equipment in the selection contract φ i Time utility should satisfy the incentive compatibility constraint. Aircraft manufacturing equipment of type i can only obtain the optimal utility when it selects a contract corresponding to its own type, as shown in formula (4):

[0037] R i -C i ≥R j -C j (4)

[0038] Among them, R i and C i The reward and training cost obtained when choosing a contract of type i for an aircraft manufacturing facility of type i; R j and C j The rewards and training costs gained when a contract of type j is selected for an aircraft manufacturing facility of type i; maximum utility is obtained when an aircraft manufacturing facility of type i selects a contract of type i.

[0039] 1.2.3: The rewards and training costs of aircraft manufacturing equipment in federated learning training should satisfy the monotonicity constraint.

[0040] Assume that the training costs of different types of aircraft manufacturing equipment in the contract meet C 0 ≤C 1 ≤…≤C i ≤…C I , the reward it receives should satisfy R 0 ≤R 1 ≤…≤R i ≤…R I .

[0041] 1.2.4: When the contract satisfies the constraints of individual rationality, incentive compatibility, and monotonicity, the expected utility of the federated learning center server is shown in Equations (5) to (7):

[0042] (IR) R i -C i ≥0 (5)

[0043] (IC) R i -C i ≥R j -C j (6)

[0044] (monotonicity) when C 0 ≤C 1 ≤…≤C i ≤…CI When R 0 ≤R 1 ≤…≤R i ≤…R I (7)

[0045] Step 1.3: On the basis of ensuring acceptable contracts for aircraft manufacturing equipment, satisfy the principle of maximizing the utility of the federated learning center server and obtain the optimal contract model.

[0046] Relax the individual rationality constraint in step 1.2.1, as shown in formula (8):

[0047] R 0 -C 0 =0 (8)

[0049] When the constraint condition of formula (9) is satisfied, all other types of aircraft manufacturing equipment satisfy the individual rationality constraint condition;

[0050] Relax the incentive compatibility constraint in step 1.2.2, as shown in formula (9):

[0051] R i -C i =R i-1 -C i-1 (9)

[0052] When the constraint condition of formula (9) is satisfied, all other types of aircraft manufacturing equipment meet the incentive compatibility constraint condition;

[0053] By relaxing the constraints in step 1.2, the optimal model of the contract is obtained, as shown in equations (10) and (11):

[0054] (IR) R 0 -C 0 =0 (10)

[0055] (IC) R i -C i =R i-1 -C i-1 (11)

[0056] The optimal solution of the contract is obtained by using equations (10) and (11), and the optimal contract is obtained.

[0057] Beneficial effects of the present invention:

[0058] The present invention firstly adopts a federated learning training framework for the training of the aircraft manufacturing equipment anomaly detection model to ensure the privacy security of data and further strengthen the privacy security of aircraft manufacturing equipment data. The aircraft manufacturing equipment anomaly detection model is trained on the basis of ensuring data privacy security, and the federated learning architecture enables the aircraft manufacturing equipment data to be trained without leaving the local device, thereby ensuring the privacy security of the data.

[0059] The present invention adopts a contract theory incentive mechanism to provide reasonable compensation for aircraft manufacturing equipment to participate in the federated learning training process, encourage aircraft manufacturing equipment to participate in the training of federated learning, and improve the accuracy of the abnormal detection model of aircraft manufacturing equipment. Based on the computing cost consumed by aircraft manufacturing equipment when training the model and the benefits gained by the central server based on the training of aircraft manufacturing equipment, an incentive mechanism model is designed to encourage aircraft manufacturing equipment to actively participate in the training of federated learning models. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] Figure 1 Create a flow chart for equipment anomaly diagnosis for aircraft based on federated learning;

[0061] Figure 2 Schematic diagram of the federated learning training process. DETAILED DESCRIPTION

[0062] The specific implementation of the present invention is further described in detail below in conjunction with the drawings and examples.

[0063] A method for abnormal diagnosis of aircraft manufacturing equipment based on federated learning, comprising the following steps:

[0064] Step 1: Establish the utility model of the federated learning center server and aircraft manufacturing equipment, satisfy the constraints of individual rationality, incentive compatibility and monotonicity, and build a contract model, which is divided into the following steps:

[0065] Step 1.1: After receiving the training task, the federated learning center server constructs the utility model of the aircraft manufacturing equipment and the federated learning center server according to the private information characteristics of the aircraft manufacturing equipment, the training cost, and the model benefit of the federated learning center server.

[0066] Step 1.1.1: The federated learning center server builds the cost model C for training the federated learning task of aircraft manufacturing equipment of type i i .

[0067] Step 1.1.2: The federated learning center server builds the utility model U for training the federated learning task of aircraft manufacturing equipment of type i f, the utility model is the reward R gained from training the federated learning task for the aircraft manufacturing equipment i Subtract the cost C required to train the task i .

[0068] U f =R i -C i (1)

[0069] Step 1.1.3: The benefit of the local model uploaded by aircraft manufacturing equipment of type i for the federated learning task is Ψ i The utility U obtained by the federated learning central server from the models uploaded by the aircraft manufacturing equipment participating in the training s The profit of its global model minus the compensation paid to each aircraft manufacturing equipment.

[0070]

[0071] Where N represents the total number of aircraft manufacturing equipment participating in the federated learning task, I represents the total number of aircraft manufacturing equipment types, and ρ i represents the prior probability of aircraft manufacturing equipment of type i, ρ i ∈[0,1].

[0072] Step 1.2: The federated learning center server ensures that the contract model meets the basic constraints acceptable to aircraft manufacturing equipment based on individual rationality constraints, incentive compatibility constraints and monotonicity constraints, and constructs a feasible contract model.

[0073] Step 1.2.1: The utility of aircraft manufacturing equipment needs to satisfy the constraints of individual rationality. The aircraft manufacturing equipment of type i chooses contract φ i When non-negative utility is obtained.

[0074] U f =R i -C i ≥0 (3)

[0075] Step 1.2.2: Aircraft manufacturing equipment in the selection contract φ i Time utility should satisfy the incentive compatibility constraint. Aircraft manufacturing equipment of type i can only obtain the optimal utility when it selects a contract corresponding to its own type, as shown in formula (4):

[0076] R i -C i ≥R j -C j (4)

[0077] Among them, R i and C i The reward and training cost obtained when selecting a contract of type i for an aircraft manufacturing facility of type i; R j and C j The rewards and training costs gained when a contract of type j is selected for an aircraft manufacturing facility of type i; when an aircraft manufacturing facility of type i selects a contract of type i, the maximum utility is obtained.

[0078] 1.2.3: The rewards and training costs of aircraft manufacturing equipment in federated learning training should satisfy the monotonicity constraint.

[0079] The training costs for different types of aircraft manufacturing equipment in the contract meet C 0 ≤C 1 ≤…≤C i ≤…C I When the reward is satisfied, 0 ≤R 1 ≤…≤R i ≤…R I .

[0080] 1.2.4: When the contract satisfies the constraints of individual rationality, incentive compatibility, and monotonicity, the expected utility of the federated learning center server is shown in Equations (5) to (7):

[0081] (IR) R i -C i ≥0 (5)

[0082] (IC) R i -C i ≥R j -C j (6)

[0083] (monotonicity) when C 0 ≤C 1 ≤…≤C i ≤…C I When R 0 ≤R 1 ≤…≤R i ≤…R I (7)

[0084] Step 1.3: On the basis of ensuring acceptable contracts for aircraft manufacturing equipment, satisfy the principle of maximizing the utility of the federated learning center server and obtain the optimal contract model.

[0085] Relax the individual rationality constraint in step 1.2.1, as shown in formula (8):

[0086] R0 -C 0 =0 (8)

[0088] When the constraint condition of formula (8) is satisfied, all other types of aircraft manufacturing equipment satisfy the individual rationality constraint condition;

[0089] Relax the incentive compatibility constraint in step 1.2.2, as shown in formula (9):

[0090] R i -C i =R i-1 -C i-1 (9)

[0091] When the constraint condition of formula (9) is satisfied, all other types of aircraft manufacturing equipment meet the incentive compatibility constraint condition;

[0092] By relaxing the constraints in step 1.2, the optimal model of the contract is obtained, as shown in equations (10) and (11):

[0093] (IR) R 0 -C 0 =0 (10)

[0094] (IC) R i -C i =R i-1 -C i-1 (11)

[0095] The optimal solution of the contract is obtained by using equations (10) and (11), and the optimal contract is obtained.

[0096] Step 2: The Federated Learning Center server signs a contract with the aircraft manufacturer;

[0097] After the Federated Learning Center server completes the construction of the optimal contract, it broadcasts the relevant tasks and the optimal contract to the aircraft manufacturing equipment. The aircraft manufacturer chooses to sign different types of optimal contracts based on the training capabilities of the equipment. The aircraft manufacturer that has signed the contract proceeds to step 3;

[0098] Step 3: The aircraft manufacturing equipment downloads the global model from the federated learning center server;

[0099] In the first round of training, the Federated Learning Center server will provide an initialization model, and the aircraft manufacturing equipment will download the initialization model for training. In subsequent rounds of training, the aircraft manufacturing equipment will download the global model updated in the previous round from the center server for training.

[0100] Step 4: The aircraft manufacturing equipment performs local training of the federated learning global model;

[0101] Step 4.1: The aircraft manufacturing equipment uses the equipment operation data generated by the local equipment to train the downloaded global model, and after the training is completed, execute step 4.2;

[0102] Step 4.2: The aircraft manufacturing equipment uploads the trained model to the federated learning center server, and uploads the signed contract to the federated learning center server.

[0103] Step 5: The federated learning center server evaluates the local model uploaded by the aircraft manufacturing equipment.

[0104] The federated learning central server will evaluate the local models uploaded by each aircraft manufacturing equipment participating in the federated learning training. If the model meets the acceptable requirements, the central server will receive the model and execute step 6. Otherwise, the model will be rejected and no corresponding compensation will be issued.

[0105] Step 6: The federated learning center server aggregates the local models uploaded by the aircraft manufacturing equipment according to the contract selection;

[0106] The federated learning center server will aggregate the federated learning model according to the selected contract for the local models uploaded by the aircraft manufacturing equipment that meets the requirements.

[0107] Step 7: The Federated Learning Center server records the corresponding remuneration of the aircraft manufacturer according to the contract;

[0108] After each round of federated learning training, the federated learning center server will update the remuneration payable to the aircraft manufacturer for participating in the federated learning global model aggregation according to the signed contract. When the phased federated learning training task is completed, the federated learning center server will issue the corresponding remuneration to the aircraft manufacturer.

[0109] Next, the federated learning center server and the aircraft manufacturing equipment conduct n (n>0) rounds of steps 3 to 7 of the federated learning training process, as shown in Figure 2 shown.

[0110] Step 8: Deploy the aircraft manufacturing equipment anomaly diagnosis model

[0111] The trained aircraft manufacturing equipment anomaly diagnosis model is deployed locally to the aircraft manufacturing equipment to detect equipment anomalies.

Claims

1. A method for abnormal diagnosis of aircraft manufacturing equipment based on federated learning, characterized in that: The following steps are involved: Step 1: After receiving the task, the federated learning center server establishes the utility model of the federated learning center server and aircraft manufacturing equipment, satisfies the constraints of individual rationality, incentive compatibility and monotonicity, and constructs a contract model; Step 2: The Federated Learning Center server signs a contract with the aircraft manufacturer; After the Federated Learning Center server completes the construction of the optimal contract, it broadcasts the relevant tasks and the optimal contract to the aircraft manufacturing equipment. The aircraft manufacturer chooses to sign different types of optimal contracts based on the training capabilities of the equipment. The aircraft manufacturer that has signed the contract proceeds to step 3. Step 3: The aircraft manufacturing equipment downloads the global model from the federated learning center server; In the first round of training, the federated learning center server will provide an initialization model, and the aircraft manufacturing equipment will download the initialization model for training. In the subsequent rounds of training, the aircraft manufacturing equipment will download the global model updated in the previous round from the center server for training. Step 4: The aircraft manufacturing equipment performs local training of the federated learning global model; Step 4.1: The aircraft manufacturing equipment uses the equipment operation data generated by the local equipment to train the downloaded global model, and after the training is completed, execute step 4.2; Step 4.2: The aircraft manufacturing equipment uploads the trained model to the federated learning center server; Step 5: The federated learning center server evaluates the local model uploaded by the aircraft manufacturing equipment; The federated learning central server will evaluate the local models uploaded by each aircraft manufacturing equipment participating in the federated learning training. If the model meets the acceptable requirements, the central server will receive the model and execute step 6. Otherwise, the central server will refuse to receive the model and will not issue the corresponding reward. Step 6: The federated learning center server aggregates the local models uploaded by the aircraft manufacturing equipment according to the contract selection; The federated learning center server will aggregate the federated learning model according to the selected contract for the local models uploaded by the aircraft manufacturing equipment that meets the requirements; Step 7: The Federated Learning Center server records the corresponding remuneration of the aircraft manufacturer according to the contract; After each round of federated learning training, the federated learning center server will update the remuneration payable to the aircraft manufacturer for participating in the federated learning global model aggregation according to the signed contract. When the phased federated learning training task is completed, the federated learning center server will issue the corresponding remuneration to the aircraft manufacturer. Next, the federated learning center server and the aircraft manufacturing equipment conduct n (n>0) rounds of steps 3 to 7 of the federated learning training process; Step 8: Deploy the aircraft manufacturing equipment anomaly diagnosis model The trained aircraft manufacturing equipment anomaly diagnosis model is deployed locally to the aircraft manufacturing equipment to detect equipment anomalies.

2. The method for abnormality diagnosis of aircraft manufacturing equipment based on federated learning according to claim 1, characterized in that: The specific process of step 1 includes the following steps: Step 1.1: After receiving the training task, the federated learning center server constructs the utility model of the aircraft manufacturing equipment and the federated learning center server according to the private information characteristics of the aircraft manufacturing equipment, the training cost, and the model income of the federated learning center server; Step 1.2: The federated learning center server ensures that the contract model meets the basic constraints acceptable to aircraft manufacturing equipment based on individual rationality constraints, incentive compatibility constraints, and monotonicity constraints, and constructs a feasible contract model; Step 1.3: On the basis of ensuring acceptable contracts for aircraft manufacturing equipment, satisfy the principle of maximizing the utility of the federated learning center server and obtain the optimal contract model.

3. The method for abnormality diagnosis of aircraft manufacturing equipment based on federated learning according to claim 1 or 2, characterized in that: The specific operation of step 1.1 in step 1 is as follows: Step 1.1.1: The federated learning center server builds the cost model C for training the federated learning task of aircraft manufacturing equipment of type i i ; Step 1.1.2: The federated learning center server builds the utility model U for training the federated learning task of aircraft manufacturing equipment of type i f , the utility model is the reward R gained from training the federated learning task for the aircraft manufacturing equipment i Subtract the cost C required to train the task i ; U f =R i -C i (1) Step 1.1.3: The benefit of the local model uploaded by aircraft manufacturing equipment of type i for the federated learning task is Ψ i ; The utility U obtained by the federated learning central server from the models uploaded by the aircraft manufacturing equipment participating in the training s The profit of its global model minus the compensation paid to each aircraft manufacturing equipment; Where N represents the total number of aircraft manufacturing equipment participating in the federated learning task, I represents the total number of aircraft manufacturing equipment types, and ρ i represents the prior probability of aircraft manufacturing equipment of type i, ρ i ∈[0,1].

4. The method for abnormality diagnosis of aircraft manufacturing equipment based on federated learning according to claim 1 or 2, characterized in that: Specific operations of step 1.2: Step 1.2.1: The utility of aircraft manufacturing equipment needs to satisfy the constraints of individual rationality. The aircraft manufacturing equipment of type i chooses contract φ i When the non-negative utility is obtained; U f =R i -C i ≥0 (3) Step 1.2.2: Aircraft manufacturing equipment in the selection contract φ i Time utility should satisfy the incentive compatibility constraint. Aircraft manufacturing equipment of type i can only obtain the optimal utility when it selects a contract corresponding to its own type, as shown in formula (4): R i -C i ≥R j -C j (4) Among them, R i and C i The reward and training cost obtained when choosing a contract of type i for an aircraft manufacturing facility of type i; R j and C j The rewards and training costs gained when a type i aircraft manufacturing facility chooses a type j contract; when a type i aircraft manufacturing facility chooses a type i contract, the maximum utility is obtained; 1.2.3: The rewards and training costs of aircraft manufacturing equipment in federated learning training should meet the monotonicity constraint; Assume that the training costs of different types of aircraft manufacturing equipment in the contract satisfy C0≤C1≤…≤C i ≤…C I , the reward it receives should satisfy R0≤R1≤…≤R i ≤…R I ; 1.2.4: When the contract satisfies the constraints of individual rationality, incentive compatibility, and monotonicity, the expected utility of the federated learning center server is shown in Equations (5) to (7): (IR) R i -W i ≥0 (5) (IC) R i -C i ≥R j -C j (6) (Monotonicity) When C0≤C1≤…≤C i ≤…C I When R0≤R1≤…≤R i ≤…R I (7).

5. The method for abnormality diagnosis of aircraft manufacturing equipment based on federated learning as claimed in claim 3, characterized in that: Specific operations of step 1.2: Step 1.2.1: The utility of aircraft manufacturing equipment needs to satisfy the constraints of individual rationality. The aircraft manufacturing equipment of type i chooses contract φ i When the non-negative utility is obtained; U f =R i -C i ≥0 (3) Step 1.2.2: Aircraft manufacturing equipment in the selection contract φ i Time utility should satisfy the incentive compatibility constraint. Aircraft manufacturing equipment of type i can only obtain the optimal utility when it selects a contract corresponding to its own type, as shown in formula (4): R i -C i ≥R j -C j (4) Among them, R i and C i The reward and training cost obtained when choosing a contract of type i for an aircraft manufacturing facility of type i; R j and C j The rewards and training costs gained when a type i aircraft manufacturing facility chooses a type j contract; when a type i aircraft manufacturing facility chooses a type i contract, the maximum utility is obtained; 1.2.3: The rewards and training costs of aircraft manufacturing equipment in federated learning training should meet the monotonicity constraint; Assume that the training costs of different types of aircraft manufacturing equipment in the contract satisfy C0≤C1≤…≤C i ≤…C I , the reward it receives should satisfy R0≤R1≤…≤R i ≤…R I ; 1.2.4: When the contract satisfies the constraints of individual rationality, incentive compatibility, and monotonicity, the expected utility of the federated learning center server is shown in Equations (5) to (7): (IR) R i -W i ≥0 (5) (IC) R i -C i ≥R j -C j (6) (Monotonicity) When C0≤C1≤…≤C i ≤…C I When R0≤R1≤…≤R i ≤…R I (7).

6. The method for abnormality diagnosis of aircraft manufacturing equipment based on federated learning according to claim 4, characterized in that: Specific operations of step 1.3: Relax the individual rationality constraint in step 1.2.1, as shown in formula (8): R0-C0=0 (8) When the constraint condition of formula (9) is satisfied, all other types of aircraft manufacturing equipment satisfy the individual rationality constraint condition; Relax the incentive compatibility constraint in step 1.2.2, as shown in formula (9): R i -C i =R i-1 -C i-1 (9) When the constraint condition of formula (9) is satisfied, all other types of aircraft manufacturing equipment meet the incentive compatibility constraint condition; By relaxing the constraints in step 1.2, the optimal model of the contract is obtained, as shown in equations (10) and (11): (IR) R0-C0=0 (10) (IC) R i -C i =R i-1 -C i-1 (11) The optimal solution of the contract is obtained by using equations (10) and (11), and the optimal contract is obtained.

7. The method for abnormality diagnosis of aircraft manufacturing equipment based on federated learning according to claim 5, characterized in that: Specific operations of step 1.3: Relax the individual rationality constraint in step 1.2.1, as shown in formula (8): R0-C0=0 (8) When the constraint condition of formula (9) is satisfied, all other types of aircraft manufacturing equipment satisfy the individual rationality constraint condition; Relax the incentive compatibility constraint in step 1.2.2, as shown in formula (9): R i -C i =R i-1 -C i-1 (9) When the constraint condition of formula (9) is satisfied, all other types of aircraft manufacturing equipment meet the incentive compatibility constraint condition; By relaxing the constraints in step 1.2, the optimal model of the contract is obtained, as shown in equations (10) and (11): (IR) R0-C0=0 (10) (IC) R i -C i =R i-1 -C i-1 (11) The optimal solution of the contract is obtained by using equations (10) and (11), and the optimal contract is obtained.

Images