669results about How to "Solve balance problems" patented technology

The invention discloses a dual-channel convolution neural network semantic segmentation method sensitive to small targets. The method comprises the following steps: a Caffe depth learning frame is used to build a non-weighted learning network and a weighted learning network; for the two-channel network, the corresponding semantic segmentation model is obtained by two-stage training. The output scoring charts of two channels are obtained by two semantic segmentation models, and the output scoring charts of two channels are fused by different model fusion algorithms, and the optimal model fusionalgorithm is selected according to the specific evaluation index. The test image is segmented according to the semantic segmentation model and the selected optimal model fusion algorithm. The invention can ensure that on the premise that the overall segmentation accuracy of the data set is better, the invention is more sensitive to the small target area existing in the image.

A semi-automatic grinder-ball grinder system for grinding ore is composed of a vibration feeder, a semi-automatic grinder with variable-speed driver and load monitor, a vibration sieve, a crusher, a sludge pump, cyclone unit with concentration, flow and granularity detections, and a ball grinder with variable-speed driver. Its control method is also disclosed.

The invention provides compound feed for teaching piglets to eat foods other than breast milk, which is used for weaning piglets and teaching the piglets to eat foods other than breast milk, and comprises the following components in part by weight: 212.4 parts of corn starch, 154.0 parts of bulked corn, 100.0 parts of 46 percent of bulked bean pulp, 40 parts of fish meal, 100 parts of wheat flour, 5 parts of calcium powder, 21 parts of calcium hydrophosphate, 2 parts of table salt, 20 parts of soybean oil, 5 parts of lysine, 2 parts of threonine, 1.5 parts of choline chloride, 10 parts of premix, 0.2 part of complex enzyme, 0.3 part of pig Duowei, 0.2 part of sweetener, 0.3 part of flavouring agent, 3 parts of baking soda, 20 parts of cane sugar, 30 parts of glucose, 50 parts of egg powder, 70 parts of soy protein concentrate, 50 parts of fermented bean pulp, 6 parts of acidizer, 100 parts of whey powder, 0.5 part of mould removing agent, 0.3 part of antioxidant and 0.3 part of mildewpreventive. In the invention, the problems of incomplete immune functions, low disease resistance, inadequate digestive ferment, incomplete upper gastrointestinal development, low digestion and absorption ability, susceptibility to diarrhea and the like of piglets.

The invention discloses a self-balanced single-wheel electric vehicle, which relates to the technical field of vehicles. A mudguard (1) is arranged on the upper side of a wheel (4); a wheel hub motor (2) is arranged in the middle of a wheel (1); a motor shaft (3) is arranged in the central hole of the wheel hub motor (2); connecting pieces (5) are arranged at two ends of the motor shaft (3); the lower side of each connection piece (5) is connected with a pedal plate (6) through a connecting rod (7); a micro processing unit (10) is connected with a three-phase PWM (Pulse Width Modulation) driving signal driving unit (15); the three-phase PWM driving signal driving unit (15) is connected with a three-phase half-bridge inverter circuit (16); and the wheel hub motor (2) is connected with a motor Hall sensor (18) and a motor current sensor (19) respectively. The self-balanced single-wheel electric vehicle has the advantages of small size, light weight, convenience for carrying, capability of solving the problem of balance due to the adoption of a balance algorithm and an advanced sensor technology, simple structure, easiness for operating, safety, reliability and low cost.

The invention discloses a preparation method of a combined phosphorus type flame-retardant acrylate pressure-sensitive adhesive. The preparation method comprises the following steps: adopting a free radical solution polymerization process and a multistep polymerization process to polymerize flame-retardant phosphorous acrylate monomers and common acrylate monomers to prepare a combined phosphorus type flame-retardant acrylate pressure-sensitive cement by using N2 as shielding gas under the conditions of normal pressure and the temperature of 70-85 DEG C; coating the pressure-sensitive cement, and drying to prepare the combined phosphorus type flame-retardant acrylate pressure-sensitive adhesive; and in order to better enhance the flame retardant property of the pressure-sensitive adhesive, optionally adding flame-retardant sizing to the pressure-sensitive cement, then coating, and drying. The method has a simple process, and the prepared pressure-sensitive adhesive has excellent properties and avoids the environmental pollution caused by a halogenated flame retardant.

The invention dsiclsoes a microgrid system with an asymmetric non-linear load and a power balancing control method. The system comprises a plurality of DG units connected in parallel and line impedors connected with all DG units. The line impedors are connected to a microgrid bus by PCC points. A three-phase balancing resistive load, an asymmetric linear load, and a diode rectifier non-linear load of a load unit are connected to the microgrid bus by PCC points. A measurement module for measuring voltage fundamental wave positive sequence and negative sequence components and harmonic wave components of the PCC points is also connected to the microgrid bus. The microgrid bus is connected with a 10-kV main power grid by a static switch and a transformer successively. According to the invention, reactive and harmonic power balancing of the microgrid is realized by using a selective virtual impedance based on the virtual fundamental positive and negative sequence impedance and the virtual variable harmonic impedance; a harmonic and unbalancing voltage compensation controller enables equal division of an unbalanced power and a harmonic power to be realized; and a problem of unbalancing of the harmonic wave and the voltage of the microgrid can be solved.

The invention relates to the network safety technology, discloses a deep message detection method, network equipment and system. The method includes: receive the data message in the data stream; determine whether the data message required to be detected or not, if yes, firstly identify the data message; if can't be identified, then copy the data message, and send one copy of the data message to the deep message detection device, that is, the DPI detection device for deep message detecting. The invention may solve the balancing problem between the DPI detection and rapid data forward performances, meet with the sensing and controlling demand of network operator for the internet services, save the network resources.

The invention provides a multichannel wireless mesh network channel distribution method which comprises the following steps: (1) according to a probability model, establishing a bidirectional network connection graph with weight, calculating a link interference degree and a node interference degree of a node in order according to link weight, and selecting a channel with a minimal node interference degree in an available channel set as a channel of a node fixed interface according to the node interference degree; (2) establishing a communication period, dividing the period into a broadcast time slot and a common data time slot, sending and receiving a broadcast message in the broadcast time slot, sending a common data packet in the common data time slot, and dividing the common data time slot into data sub time slots, wherein different sub time slots can use different channels to communicate with different nodes. According to the invention, a plurality of nonoverlap channels are fully utilized, network interference is effectively reduced, and network performances such as a network throughput, an end-to-end time delay and the like are raised.

The invention discloses a knowledge map representation learning method which combines an entity and a relationship description. Text description information of the entity and the relationship is takeninto consideration, two information sources of structured information and the text description information of a triple are well integrated, based on combined extracts of an end-to-end model of neuralnetworks conducted on the entity and the relationship, a balance factor is set to balance the structured information and the text description information, and different score functions are defined according to different prediction objects; a loss function is used for correlating an entity vector and a relationship vector, the loss function is optimized, and when an optimization goal is reached, the vectors of each entity and relationship in the knowledge map and the text description information can be learned. The method solves the sparsity and imbalance of entity and relationship in a knowledge base, accurately and effectively represents the interconnection between the entity and relationship, applies the interconnection to the large-scale knowledge map, and has good practicality.

The invention relates to a software defect predication method based on cost sensitivity and semi-supervision. The software defect predication method comprises the following steps: S1, collecting a source code file of software to be predicated through a version control tool; S2, extracting a measurement element value from the source code file, wherein the source code file obtained by S1 is composed of I modules; S3, obtaining a sampling result set through a measurement element value selecting module in a sampling manner; S4, constructing a training set by a labeled sampling result set and a label-free sampling result set; S5, extracting a target function and solving a classification function enabling the target function to be minimum; and S6, predicating a module to be predicated through the classification function and outputting a predicated result. The method provided by the invention fuses semi-supervision and cost sensitivity thoughts to construct a software defect predication model and two problems in software defect predication that defect data is difficult to obtain and types are not balanced are solved; and the accuracy of the predicated result is extremely improved.

The invention relates to a fretting wear test device, in particular to a high temperature and high pressure in-situ fretting wear test device, and solves the problems that in-situ fretting wear is difficult to realize at a high temperature and a high pressure; fretting amplitude and frequency are difficult to control accurately; electrochemical signals are hard to measure during fretting wear. The fretting wear test device is provided with a vibration exciter, a thermocouple, a working electrode / auxiliary electrode seat, an autoclave cover, a hydraulic cylinder, an autoclave body, an upright, a positioning plate, a first sample holding and loading device, a moving shaft, an upper linear guideway, a working electrode, a moving shaft guide bracket, an auxiliary electrode, a connecting nut I, a reference electrode seat, a reference electrode, a lower linear guideway, an upper guide platform plate, an upper guide rod, a lower guide rod, a lower guide platform plate and a connecting nut II. The device is capable of performing in-situ fretting wear on the surface of a sample at a high temperature and a high pressure. The reference electrode, the working electrode and the auxiliary electrode mounted on an autoclave work with an electrochemical workstation to detect the electrochemical signals during fretting wear.

The invention discloses a dynamic unbalance online testing and fault identification method and system for wind turbine generators. The method includes the following steps that within a set time duration, synchronous sampling on the generator power, the pitch angle position, the wind wheel rotating speed and the accelerated speed of a cabin in the front and back direction is carried out; according to the generator power and the pitch angle position collected according to synchronous sampling, whether a draught fan runs under the designed testing work condition or not is judged; if it is judged that the draught fan runs under the designed testing work condition, at least spectral analysis is carried out on time series of the collected wind wheel rotating speed; according to results of the spectral analysis on the time series of the collected wind wheel rotating speed, whether the draught fan has impeller unbalance faults or not is judged; if the draught fan has impeller unbalance faults, either pneumatic imbalance or quality moment imbalance is the reason of unbalancedness is judged according to the results of the spectral analysis on the time series of the accelerated speed of the cabin in the front and back direction. The invention further discloses a corresponding system.

The invention discloses a knowledge-graph representation learning method based on multiple semantemes. The method includes the steps that different semantemes of relationships are considered, a modelbased on translation between physical vectors and relationship vectors and different semantemes of relationships in a triple structure are used in cooperation with projected vectors of head-tail entities to define relationship matrixes, correlation between the physical vectors and the relationship vectors in the triple structure is well expressed, and more parameters are not introduced; the physical vectors and the relationship vectors are correlated through loss functions, and the loss functions are optimized; when the optimization target is achieved, all the physical vectors and all the relationship vectors in knowledge graphs can be learned. By means of the knowledge-graph representation learning method based on the multiple semantemes, the problems of the heterogeneity and the unbalancedness between entities and relationships in a knowledge base are solved, and the entities and the relationships and correlation between the entities and the relationships are more accurately expressed, and are applied to large-scale knowledge graphs.

An intelligent cross-language speech recognition and conversion method claimed by the present invention classifies speech data based on the language family, establishes inter-language distances, initially determines a language family to which to-be-recognized speech data belongs, and then performs language refined identification in the language family. After the first language identification error, neighbor language families can be further found based on established inter-language distances, and the language family is established; and after the language is identified, the language is convertedinto a standardized text, then the text is subjected to word segmentation, word frequency statistics and other processing to establish the mapping relationship, which facilitates subsequent voice query. On the one hand, the method effectively solves the disadvantages that the efficiency and the speed cannot be balanced in the current speech recognition. On the other hand, the processing of the speech text conversion is more reasonable, and the establishment of mapping relationship makes the recognition conversion efficiency more accurate.

The invention relates to a multifunctional road detection device and a test method for tires. Two suspension systems and two tire steering mechanisms of the multifunctional road detection device are arranged on a frame and are respectively independently connected with the tested tires on two sides of the multifunctional road detection device, so that the tires can be independently steered and can be adjusted in internally gathered or externally scattered states, and lateral inclination angles of the tires can be adjusted. The multifunctional road detection device and the test method have the advantages that data of rotation angles, lateral force, longitudinal acceleration, lateral acceleration and the like of the tires can be acquired by various relevant sensors during driving, so that mechanical properties of the tires under on-site road conditions can be detected; the problem of difficulty in simulating actual service road surfaces by existing indoor detection devices for tires and problems of unbalanced force under lateral deviation and lateral rolling working conditions and single detection working conditions of existing outdoor detection devices for the tires can be solved; the multifunctional road detection device is simple in structure and easy to control and adjust.

The invention discloses a mechanical-arm hand object grabbing method based on depth learning, and belongs to the technical field of multi-degree-of-freedom manipulator control. Before object grabbing,labels are made in advance by adopting a depth learning network structure, training classification is conducted, and a database is made with one label corresponding to one grabbing angle; and when auser says an instruction, a to-be-grabbed object is determined through speech recognition, then the object is found through image recognition and positioning, an image coordinate and the mechanical-arm hand grabbing angle are returned, a BP neural network subjected to particle swarm optimization corrects the image coordinate, finally, the rotating angles required by all motors are resolved reversely through a GRNN network, and grabbing of a manipulator is finished after rotating to the target. According to the mechanical-arm hand object grabbing method based on depth learning, grabbing of theselected target can be achieved, and meanwhile, the problem of instable grabbing is avoided.

The invention provides an edible blending oil with balanced nutrients and a production method thereof. The proportion of saturated fatty acid to monounsaturated fatty acid to polyunsaturated fatty acids of the blending oil is designed based on daily intake animal meat fat, so that the standard of and 1t; 1:1:1, omega-6 and omega-3 of 4-5:1) recommended by Chinese Society of Nutrition for the final intake of dietary fatty acid of people is met. The production method comprises the steps of accurately weighting various kinds of vegetable oil, vitamin E equivalent to 0.5 percent of the total mass of the vegetable oil in mass and / or tea polyphenol equivalent to 0.3 percent of the total mass of the vegetable oil in mass according to different formulas, adding the weighted materials into a dosing tank, keeping the temperature to be 20-40 DEG C, evenly stirring the mixture for 20-40 minutes at low speed, filtering the mixture and filling the obtained product to obtain the finished product, i.e. the edible blending oil with the balanced nutrients.

The invention provides a bean pulp rainbow trout parr compound feed and a preparation method thereof. Every 100kg of bean pulp rainbow trout parr compound feed is manufactured by 10 to 15kg of steam fish meal, 45 to 55kg of full fat expanded soybean, 10 to 15kg of corn protein powder, 2 to 5kg of isolate soya protein, 3 to 5kg of beer yeast, 0.5 to 1.0kg of Squid paste, 0.5 to 1.0kg of betaine, 3 to 6kg fo flour, 1.5 to 2.5kg of soybean lecithin, 2 to 3kg of fish oil, 2 to 3kg of soybean oil, 0.3 to 0.5kg of methionine, 0.1 to 0.3kg of lysine, 0.3 to 0.5kg of Choline Chloride, 0.5 to 1.0kg of composite vitamin, 0.5 to 1.0kg of composite mineral salt and water. The invention aims at the digesting physiological characteristic during the parr period of the rainbow trout and solves the balancing problem of amino acid by optimizing and adjusting the prescription of the feed, and solves the problems of anti-nutritional factors and palatability by comprehensively using enzyme preparation and feeding promoting agent. The feed of the invention has the advantages of remarkably promoting the quick growth of the rainbow trout parr, improving the surviving rate of the rainbow trout parr, improving the digesting and absorbing rates of the rainbow trout parr on the feed, reducing the drainage of P and N, reducing the culturing cost, and the like.

The present invention provides a multi-microgrid flexible interconnection system and a control method thereof. The system includes a plurality of microgrids and multi-port flexible interconnectors, wherein the multi-port flexible interconnectors are arranged between the feeder lines of the plurality of microgrids; and the multi-port flexible interconnectors are connected into the tail ends of the adjacent feeder lines of the plurality of microgrids, so that the flexible interconnection of the plurality of microgrids is realized. According to the multi-microgrid flexible interconnection system and the control method thereof of the invention, the multi-port flexible interconnectors, replacing traditional contact switches, are connected into the tail ends of the adjacent feeder lines between the different microgrids, so that the flexible interconnection of the plurality of microgrids is realized; due to the controllability of power electronic devices, power flow interaction among the plurality of microgrids can be realized. With the multi-microgrid flexible interconnection system and the control method thereof of the invention, multi-microgrid loop network operation can be realized with no short-circuit current increased; the local absorbing capacity of new energy resources can be improved; load balance between the feeder lines can be improved; mutual backup and mutual power supply of the adjacent microgrids can be realized; line loss can be decreased; the reliability of power supply can be improved; and voltage quality at the tail ends of the feeder lines can be maintained.

During the production of two-way extended polypropylene film, the tube extruder extruded polypropylene tube blank is cooled, set in a preheating extending tower and a high temperature setting tower, and wound in a winding system. The special equipment includes blank tube extruder, preheating extending tower, and corona film edge cutting and winding system as well as high temperature setting towercomprising film bubble stabilizing A-shaped plates, heating mechanism and cooling mechanism. The present invention makes it possible to make both single-layered film and multi-layered film, especially superlow temperature blast material with high quality, no surface damage and good longitudinal performance.

The invention relates to a one-step method dihydrate-semi-hydrated wet production process for phosphoric acid. The one-step method dihydrate-semi-hydrated wet production process comprises the following steps: carrying out leaching reaction and dehydrate-semi-hydrated conversion; growing grains; separating, filtering and washing; by controlling technological indicators of all the steps, finished product phosphoric acid filtrate with high concentration and by-product phosphogypsum with high quality are obtained. The finished product phosphoric acid filtrate obtained by production of the one-stepmethod dihydrate-semi-hydrated wet production process disclosed by the invention has the advantages that the concentration of P2O5 is about 35 percent, concentration input of the phosphoric acid canbe avoided and an energy-saving effect is good. In addition, the phosphorus conversion rate and the washing rate are high; the final by-product phosphogypsum has the characteristics that the total content of the P2O5 is smaller than or equal to 0.5 percent, the content of water-soluble P2O5 is smaller than or equal to 0.15 percent and the yield of the P2O5 reaches 98 percent. The one-step method dihydrate-semi-hydrated wet production process disclosed by the invention retains the characteristics of good energy-saving effect, high phosphorus yield and high quality of byproduct alpha-semi-hydrated gypsum of a traditional two-step method dehydrate-semi-hydrated technology; meanwhile, a dihydrate filtration procedure is omitted, the process flow is shortened and the operation rate of a deviceis improved, so that the problem that water balance of a system cannot be realized due to overhigh content of wash water for filter cloth is solved.

The invention relates to the technical field of robot path planning, in particular to a multi-robot path planning method based on priori knowledge and a DQN algorithm, and the method comprises the steps: initializing parameters of a multi-robot system; judging whether a special state occurs or not, if so, selecting an action instruction corresponding to the maximum prior of the prior Q value vector, and if not, generating an action instruction by an epsilon-greed strategy; then, calculating and generating running state parameters and a reward function after the robot executes the action instruction, storing related data in an experience pool, and updating a target Q network and according to the target Q network and the initial state parameters of the multi-robot system, repeatedly executing the selection action instruction and generating the state parameters to plan an optimal path of the multi-robot system. The problems that when the DQN algorithm is used for path planning of the multi-robot system, the convergence rate of the target Q network is low, and the training randomness is too high can be better solved.

The invention discloses a joint optimization method of safety capacity and energy consumption in a wireless relay network, and relates to the field of wireless relay network cooperative communication. The method comprises the steps as follows: a safety capacity value from a source node to a destination node is restrained to be SCcon; a minimum energy consumption function of an interference node is obtained by using a beam forming technology according to the SCcon; a minimum energy consumption function of a relay node is obtained according to the SCcon; a minimum energy consumption function Dpower of a system is obtained; a minimum energy consumption value of the system is obtained at an extreme point of a minimum value of the Dpower, and the restrained safety capacity is released; and an efficiency function RSE of the system energy consumption is calculated. The invention provides a joint optimization strategy of the relay node and the interference node, and an accurate expression of equivalent safety capacity maximization corresponding to the unit energy consumption of the system is obtained, so that each part of energy reaches the maximal usage value, and the problem of balance between the safety and the energy efficiency in the wireless relay network under a single sniffing node is effectively solved.

The invention discloses a network encrypted traffic recognition method and device. The method comprises a preprocessing stage and a classification stage. In the preprocessing stage, original flow is subjected to flow segmentation, sampling, vectorization and standardization, a sampling scheme in large flow is provided, and the classification problem of the large flow is solved. In the classification stage, spatial feature capture and abstract feature extraction are performed by using a CNN, and then traffic time sequence features are learned by using stacked bidirectional LSTM on the basis ofabstract features so that automatic feature extraction and efficient recognition of encrypted traffic can be realized. The method has universality, can automatically extract the space-time features ofthe encrypted traffic without manual feature design of experts, and can adapt to traffic feature changes caused by different encryption technologies and confusion technologies.

The present invention discloses a remote fire detection system and method based on a video image. The system comprises: an image collection module configured to collect infrared images through a dual-band video collection technology and transmit the infrared images to a control module through a video decoder according to the features of the imaging of an infrared camera; the control module configured to achieve feature extraction for the infrared images through a DSP, perform flame recognition for the extracted image features through a recognition algorithm and upload a recognition result to aremote control center; a network communication module, wherein the control module is connected with the remote control center through a physical layer transceiver and a network interface in the network communication module, if fire occurs, the remote control center issues a control command to achieve an Ethernet communication function through the network communication module; and a wireless remote alarm module configured to achieve wireless network remote alarm. The remote fire detection system and method based on the video image improve the accuracy, the sensitivity and the reliability of the flame recognition algorithm and provide an effective method for large-space remote fire detection.

The invention provides a bean pulp rainbow trout parr compound feed and a preparation method thereof. Every 100kg of bean pulp rainbow trout parr compound feed is manufactured by 6 to 12kg of steam fish meal, 50 to 60kg of expanded bean pulp, 10 to 15kg of corn protein powder, 3 to 5kg of beer yeast, 0.5 to 1.0kg of Squid paste, 0.5 to 1.0kg of betaine, 2 to 6kg fo flour, 1.5 to 2.5kg of soybean lecithin, 5 to 6kg of fish oil, 5 to 63kg of soybean oil, 0.3 to 0.5kg of methionine, 0.1 to 0.3kg of lysine, 0.3 to 0.5kg of Choline Chloride, 0.5 to 1.0kg of composite vitamin, 0.5 to 1.0kg of composite mineral salt and water. The invention aims at the digesting physiological characteristic during the parr period of the rainbow trout and solves the balancing problem of amino acid by optimizing and adjusting the prescription of the feed, and solves the problems of anti-nutritional factors and palatability by comprehensively using enzyme preparation and feeding promoting agent. The invention selects the feed materials with high biological availability, and has the advantages of remarkably promoting the quick growth of the rainbow trout parr, improving the surviving rate of culturing, improving the digesting and absorbing rates of the rainbow trout parr on the feed, reducing the drainage of P and N, reducing the culturing cost, and the like. The materials used by the compound feed of the invention have the advantages of reliable source, stable quality and low cost.

This invention discloses a kind of high-efficiency compensating three-phase imbalanced load and reactive power apparatus. It concludes: three-phase current of three-phase electric system, three-phase voltage sampling circuit, power controller and different three-phase compensating elements controlled by the power controller. According to the sampling results, the power controller controls each unidirectional switch of three-phase compensating elements of different topological structure respectively. It makes them can either compensate together or compensate solely; it can either compensate reaction or balance action; it can either use for static compensating or use for dynamic compensating; it can either adjust three-phase current of adjust three-phase imbalanced voltage. It can be used for improving and replacing the reactive compensation apparatus of the three-phase-four-line mode imbalanced load electric system.

The invention provides a high linearity LINC transmitter. The LINC transmitter comprises a digital signal processing unit which is used for accomplishing the separation calculation of baseband signals and the self-adapting rectification calculation of feedback signals, an analog signal processing unit which is used for converting the digital baseband signals into analog baseband signals and transmitting the analog baseband signals into an emission and frequency carrying part by quadrature modulation, an amplifying unit which is used for accomplishing the amplification of power of radio-frequency signals, a synthesis unit which is used for synthesizing and outputting the two channels of signals, and a feedback unit which is used for accomplishing the operation of collecting the radio-frequency signals, generating intermediate frequency signals by down-conversion and transmitting the intermediate frequency signals into a baseband so as to calibrate the imbalance of amplitude and phase positions in a system.

The invention discloses a multifunctional engraving, punching and cutting all-in-one machine, which comprises a framework. A pulley capable of sliding back and forth is mounted on the framework, a head capable of moving left and right is mounted on the pulley, and a punching device, a plasma cutting device and an engraving device are mounted on the head. The multifunctional engraving, punching and cutting all-in-one machine not only has functions of an engraving machine, but also has functions of a punching machine and functions of a plasma cutting machine, and is fast in machining speed, production efficiency is greatly improved, machining precision is enhanced, and production cost is saved.

The invention discloses a method for processing thin wall welder cases of revolving body structures, belonging to the technical field of aero-engine metal plate forming and welding. According to the invention, a whole pre-deformation welding / forming method is used for solving the problems that sizes of mounting edges are out of tolerance, coaxiality of front and rear mounting edges with main bodies after welding is out of tolerance and residual stress distribution of welding components is unbalanced. The method provided by the invention comprises the following steps of: carrying out technological structure welding deformation allowance compensation design on a mounting edge and a main body of a thin wall welder case before the welding according to the plastic deformation characteristics of materials; plastically forming an expansion block according to an expander work part, size and final structural size requirement of parts; carrying out automatic argon arc welding on the mounting edge and the main body; assembling the welded welder case and expansion block on the expander work part to carry out expansion plastic forming and controlling the expansion forming size through adjusting pressure of the expander; and unloading the pressure of the expander, recovering the expander work part and unloading the parts to complete the welding / forming process.